Simple cells in the striate cortex have been depicted as half-wave-rectified linear operators. Complex cells have been depicted as energy mechanisms, constructed from the squared sum of the outputs of quadrature pairs of linear operators. However, the linear/energy model falls short of a complete explanation of striate cell responses. In this paper, a modified version of the linear/energy model is presented in which striate cells mutually inhibit one another, effectively normalizing their responses with respect to stimulus contrast. This paper reviews experimental measurements of striate cell responses, and shows that the new model explains a significantly larger body of physiological data.

Absorbance spectra were recorded by microspectrophotometry from 39 different rod and cone types representing amphibians, reptiles, and fishes, with A1- or A2-based visual pigments and λmax ranging from 357 to 620 nm. The purpose was to investigate accuracy limits of putative universal templates for visual pigment absorbance spectra, and if possible to amend the templates to overcome the limitations. It was found that (1) the absorbance spectrum of frog rhodopsin extract very precisely parallels that of rod outer segments from the same individual, with only a slight hypsochromic shift in λmax, hence templates based on extracts are valid for absorbance in situ; (2) a template based on the bovine rhodopsin extract data of Partridge and De Grip (1991) describes the absorbance of amphibian rod outer segments excellently, contrary to recent electrophysiological results; (3) the λmax/λ invariance of spectral shape fails for A1 pigments with small λmax and for A2 pigments with large λmax, but the deviations are systematic and can be readily incorporated into, for example, the Lamb (1995) template. We thus propose modified templates for the main “α-band” of A1 and A2 pigments and show that these describe both absorbance and spectral sensitivities of photoreceptors over the whole range of λmax. Subtraction of the α-band from the full absorbance spectrum leaves a “β-band” described by a λmax-dependent Gaussian. We conclude that the idea of universal templates (one for A1- and one for A2-based visual pigments) remains valid and useful at the present level of accuracy of data on photoreceptor absorbance and sensitivity. The sum of our expressions for the α- and β-band gives a good description for visual pigment spectra with λmax > 350 nm.

We have previously documented the exquisite motion sensitivity of neurons in extrastriate area MT by studying the relationship between their responses and the direction and strength of visual motion signals delivered to their receptive fields. These results suggested that MT neurons might provide the signals supporting behavioral choice in visual discrimination tasks. To approach this question from another direction, we have now studied the relationship between the discharge of MT neurons and behavioral choice, independently of the effects of visual stimulation. We found that trial-to-trial variability in neuronal signals was correlated with the choices the monkey made. Therefore, when a directionally selective neuron in area MT fires more vigorously, the monkey is more likely to make a decision in favor of the preferred direction of the cell. The magnitude of the relationship was modest, on average, but was highly significant across a sample of 299 cells from four monkeys. The relationship was present for all stimuli (including those without a net motion signal), and for all but the weakest responses. The relationship was reduced or eliminated when the demands of the task were changed so that the directional signal carried by the cell was less informative. The relationship was evident within 50 ms of response onset, and persisted throughout the stimulus presentation. On average, neurons that were more sensitive to weak motion signals had a stronger relationship to behavior than those that were less sensitive. These observations are consistent with the idea that neuronal signals in MT are used by the monkey to determine the direction of stimulus motion. The modest relationship between behavioral choice and the discharge of any one neuron, and the prevalence of the relationship across the population, make it likely that signals from many neurons are pooled to form the data on which behavioral choices are based.

Dynamic random-dot stimuli have been widely used to explore central mechanisms of motion processing. We have measured the responses of neurons in area MT of the alert monkey while we varied the strength and direction of the motion signal in such displays. The strength of motion is controlled by the proportion of spatiotemporally correlated dots, which we term the correlation of the stimulus. For many MT cells, responses varied approximately linearly with stimulus correlation. When they occurred, nonlinearities were equally likely to be either positively or negatively accelerated. We also explored the relationship between response magnitude and response variance for these cells and found, in general agreement with other investigators, that this relationship conforms to a power law with an exponent slightly greater than 1. The variance of the cells' discharge is little influenced by the trial-to-trial fluctuations inherent in our stochastic display, and is therefore likely to be of neural origin. Linear responses to these stochastic motion stimuli are predicted by simple, low-level motion models incorporating sensors having relatively broad spatial and temporal frequency tuning.

A quantitative model is developed to predict the perceived direction of moving two-dimensional patterns. The model incorporates both a simple motion energy pathway and a “texture boundary motion” pathway that incorporates response squaring before the extraction of motion energy. These pathways correspond to Fourier and non-Fourier motion pathways and are hypothesized to reflect processing in the VI-MT and V1-V2-MT pathway, respectively. A cosine-weighted sum of these pathways followed by competitive feedback inhibition accurately predicts the perceived direction for patterns composed of two cosine gratings at different orientations (“plaids”). The model also predicts direction discrimination, differences between foveal and peripheral viewing, changes in perceived direction with exposure duration, motion masking, and motion transparency.

Mechanisms supporting orientation selectivity of cat striate cortical cells were studied by stimulation with two superimposed sine-wave gratings of different orientations. One grating (base) generated a discharge of known amplitude which could be modified by the second grating (mask). Masks presented at nonoptimal orientations usually reduced the base-generated response, but the degree of reduction varied widely between cells. Cells with narrow orientation tuning tended to be more susceptible to mask presence than broadly tuned cells; similarly, simple cells generally showed more response reduction than did complex cells.

The base and mask stimuli were drifted at different temporal frequencies which, in simple cells, permitted the identification of individual response components from each stimulus. This revealed that the reduction of the base response by the mask usually did not vary regularly with mask orientation, although response facilitation from the mask was orientation selective. In some sharply tuned simple cells, response reduction had clear local maxima near the limits of the cell's orientation-tuning function.

Response reduction resulted from a nearly pure rightward shift of the response versus log contrast function. The lowest mask contrast yielding reduction was within 0.1–0.3 log unit of the lowest contrast effective for excitation.

The temporal-frequency bandpass of the response-reduction mechanism resembled that of most cortical cells. The spatial-frequency bandpass was much broader than is typical for single cortical cells, spanning essentially the entire visual range of the cat.

These findings are compatible with a model in which weak intrinsic orientation-selective excitation is enhanced in two stages: (1) control of threshold by nonorientation-selective inhibition that is continuously dependent on stimulus contrast; and (2) in the more narrowly tuned cells, orientation-selective inhibition that has local maxima serving to increase the slope of the orientation-tuning function.

The responses of simple cells were recorded from the visual cortex of cats, as a function of the position and contrast of counterphase and drifting grating patterns, to assess whether direction selectivity can be accounted for on the basis of linear summation. The expected responses to a counterphase grating, given a strictly linear model, would be the sum of the responses to the two drifting components. The measured responses were not consistent with the linear prediction. For example, nearly all cells showed two positions where the responses approached zero (i.e. two “null phase positions”); this was true, even for the most direction selective cells. However, the measured responses were consistent with the hypothesis that direction selectivity is a consequence of the linear spatiotemporal receptive-field structure, coupled with the nonlinearities revealed by the contrast-response function: contrast gain control, halfwave rectification, and expansive exponent. When arranged in a particular sequence, each of these linear and nonlinear mechanisms performs a useful function in a general model of simple cells. The linear spatiotemporal receptive field initiates stimulus selectivity (for direction, orientation, spatial frequency, etc.). The expansive response exponent enhances selectivity. The contrast-set gain control maintains selectivity (over a wide range of contrasts, in spite of the limited dynamic response range and steep slope of the contrast-response function). Rectification conserves metabolic energy.

The glaucomas are neurodegenerative diseases involving death of retinal ganglion cells and optic nerve head excavation. A major risk factor for this neurodegeneration is a harmfully elevated intraocular pressure (IOP). Human glaucomas are typically complex, progressive diseases that are prevalent in the elderly. Family history and genetic factors are clearly important in human glaucoma. Mouse studies have proven helpful for investigating the genetic and mechanistic basis of complex diseases. We previously reported inherited, age-related progressive glaucoma in DBA/2J mice. Here, we report our updated findings from studying the disease in a large number of DBA/2J mice. The period when mice have elevated IOP extends from 6 months to 16 months, with 8–9 months representing an important transition to high IOP for many mice. Optic nerve degeneration follows IOP elevation, with the majority of optic nerves being severely damaged by 12 months of age. This information should help with the design of experiments, and we present the data in a manner that will be useful for future studies of retinal ganglion cell degeneration and optic neuropathy.

Slow horizontal head and body rotation occurs in mice and rats when the visual field is rotated around them, and these optomotor movements can be produced reliably in a virtual-reality system. If one eye is closed, only motion in the temporal-to-nasal direction for the contralateral eye evokes the tracking response. When the maximal spatial frequency capable of driving the response (“acuity”) was measured under monocular and binocular viewing conditions, the monocular acuity was identical to the binocular acuity measured with the same rotation direction. Thus, the visual capabilities of each eye can be measured under binocular conditions simply by changing the direction of rotation. Lesions of the visual cortex had no effect on the acuities measured with the virtual optokinetic system, whereas perceptual thresholds obtained previously with the Visual Water Task are. The optokinetic acuities were also consistently lower than acuity estimates from the Visual Water Task, but contrast sensitivities were the same or better. These data show that head-tracking in a virtual optokinetic drum is driven by subcortical, lower frequency, and contralateral pathways.

Existing models of the primate photopic electroretinogram (ERG) attribute the light-adapted b–wave to activity of depolarizing bipolar cells (DBCs), mediated through a release of potassium that is monitored by Müller cells. However, possible ERG contributions from OFF-bipolar cells (HBCs) and horizontal cells (HzCs) have not been explored. We examined the contribution of these hyperpolarizing second-order retinal cells to the photopic ERG of monkey by applying glutamate analogs to suppress photoreceptor transmission selectively to HBC/HzCs vs. DBCs.

ERGs of Macaca monkeys were recorded at the cornea before and after intravitreal injection of drugs. Photopic responses were elicited by bright 200–220 ms flashes on a steady background of 3.3 log scotopic troland to suppress rod ERG components.

2–amino-4–phosphonobutyric acid (APB), which blocks DBC light responses, abolished the photopic b–wave and indicated that DBC activity is requisite for photopic b–wave production.

However, applying cis–2,3–piperidine dicarboxylic acid (PDA) and kynurenic acid (KYN), to suppress HBCs/HzCs and third-order neurons, revealed a novel ERG response that was entirely positive and was sustained for the duration of the flash. The normally phasic b–wave was subsumed into this new response. Applying n–methyl-dl-aspartate (NMA) did not replicate the PDA+KYN effect, indicating that third-order retinal cells are not involved. This suggests that HBC/HzC activity is critical for shaping the phasic b–wave.

Components attributable to depolarizing vs. hyperpolarizing cells were separated by subtracting waveforms after each drug from responses immediately before. This analysis indicated that DBCs and HBC/HzCs each can produce large but opposing field potentials that nearly cancel and that normally leave only the residual phasic b–wave response in the photopic ERG.

Latency of the DBC component was 5–9 ms slower than the HBC/HzC component. However, once activated, the DBC component had a steeper slope. This resembles properties known for the two types of cone synapses in lower species, in which the sign-preserving HBC/HzC synapse has faster kinetics but probably lower gain than the slower sign-inverting G-protein coupled DBC synapse.

A human patient with “unilateral cone dystrophy” was found to have a positive and sustained ERG that mimicked the monkey ERG after PDA+KYN, indicating that these novel positive photopic responses can occur naturally even without drug application.

These results demonstrate that hyperpolarizing second-order neurons are important for the primate photopic ERG. A “Push-Pull Model” is proposed in which DBC activity is requisite for b–wave production but in which HBC/HzC activity limits the amplitude and controls the shape of the primate photopic b–wave.

Retinal horizontal cells of four rodent species, rat, mouse, gerbil, and guinea pig were examined to determine whether they conform to the basic pattern of two horizontal cell types found in other mammalian orders. Intracellular injections of Lucifer-Yellow were made to reveal the morphologies of individual cells. Immunocytochemistry with antisera against the calcium-binding proteins calbindin D-28k and parvalbumin was used to assess population densities and mosaics.

Lucifer-Yellow injections showed axonless A-type and axon-bearing B-type horizontal cells in guinea pig, but revealed only B-type cells in rat and gerbil retinae. Calbindin immunocytochemistry labeled the A-and B-type populations in guinea pig, but only a homogeneous regular mosaic of cells with B-type features in rat, mouse, and gerbil. All calbindin-immunoreactive horizontal cells in the latter species were also parvalbumin-immunoreactive; comparison with Nissl-stained retinae showed that both antisera label all of the horizontal cells. Taken together, the data from cell injections and the population studies provide strong evidence that rat, mouse, and gerbil retinae have only one type of horizontal cell, the axon-bearing B-type, where as the guinea pig has both A-and B-type cells. Thus, at least three members of the family Muridae differ from other rodents and deviate from the proposed mammalian scheme of horizontal cell types.

The absence of A-type cells is apparently not linked to any peculiarities in the photoreceptor populations, and there is no consistent match between the topographic distributions of the horizontal cells and those of the cone photoreceptors or ganglion cells across the four rodent species. However, the cone to horizontal cell ratio is rather similar in the species with and without A-type cells.

The functions of the primate color-opponent and broad-band channels were assessed by examining the visual capacities of rhesus monkeys following selective lesions of parvocellular and magnocellular lateral geniculate nucleus, which respectively relay these two channels to the cortex. Parvocellular lesions impaired color vision, high spatial-frequency form vision, and fine stereopsis. Magnocellular lesions impaired high temporal- frequency flicker and motion perception but produced no deficits in stereopsis. Low spatial-frequency form vision, stereopsis, and brightness perception were unaffected by either lesion. Much as the rods and cones of the retina can be thought of as extending the range of vision in the intensity domain, we propose that the color-opponent channel extends visual capacities in the wavelength and spatial-frequency domains whereas the broad-band channel extends them in the temporal domain.

Latencies to small flashing spots of light were measured in different layers of areas V1 and V2 in anesthetized and paralyzed macaque monkeys. The shortest latencies were found in layers 4Cα and 4B of area V1. Latencies in layer 4Cβ were on average 20 ms longer than those in 4Cα and 4B. The shortest latencies in area V2 were observed in the infragranular layers and they did not differ significantly from those found in the infragranular layers in V1. Similarly, latencies in the supragranular layers of V2 were not significantly different from those measured in the supragranular layers of V1. These results show that, in area V1, neurons of the magnocellular pathway are activated on average 20 ms earlier than those of the parvocellular pathway. Our data also suggest that much processing begins simultaneously in areas V1 and V2.

Human, macaque monkey, and rat retinas were immunostained with a polyclonal antibody preparation against purified recoverin, a 23-kD calcium-binding protein isolated from bovine retina that localizes to rods and cones (Dizhoor et al., 1991). In addition to immunoreactive photoreceptors, we have identified subpopulations of recoverin-positive bipolar cells in all three species. Results from immunostaining with progressive dilutions of anti-recoverin and preadsorption of the antibody with a dilution series of purified recoverin showed that photoreceptors and bipolar cells had similar affinities for the antibody and suggested that the molecule recognized by the antibody in both cell types is recoverin. Immunoreactivity for recoverin and protein kinase C, a selective marker for all rod bipolar cells, was found in separate bipolar cell populations. Recoverin immunoreactivity is therefore a characteristic of certain cone bipolar cell types.

In rat retina, anti-recoverin labeled two morphologically distinct subpopulations of cone bipolar cells whose axonal arbors stratified at different depths in the inner plexiform layer (IPL). The bipolar cells labeled with anti-recoverin did not correspond to those that were reactive for calbindin, another cone bipolar cell marker.

Human and monkey retinas also had two populations of cone bipolar cells that were recoverin-positive. One population showed a distinct pattern of narrow bistratification at the outer border of the IPL and a regular mosaic arrangement of its axonal arbors, suggesting that the entire population of a single cone bipolar type was labeled. Cell density, dendritic morphology, and axonal-field size and stratification indicate that anti-recoverin selectively stains the flat midget (presumed OFF-center) cone bipolar cell type observed previously in Golgi preparations. By contrast the second bipolar cell population had axonal stratification in the inner half of the IPL and showed an unusual but consistent morphology and spatial distribution. Individual cells were intensely stained but were present at an extremely low density (~2−5 cells/mm2). These cells had multibranched dendritic trees characteristic of the diffuse bipolar cell class, but very small axonal fields in the size range of the midget bipolar class. Neither of the two recoverin-positive bipolar cell types in monkey was labeled with anti-calbindin or anti-cholecystokinin. An antibody preparation against bovine pineal hydroxyindole-O-methyltransferase (HIOMT) labeled photoreceptors and bipolar cells that closely resembled the recoverin-positive bipolar cells in human and rat retinas. Preadsorption of this antibody preparation with purified recoverin abolished immunostaining of the bipolar cells, suggesting that the anti-HIOMT preparation contains antibodies against recoverin, which is known to be present in the bovine pineal gland.

A number of lines of evidence suggest that, in the macaque monkey, inferior parietal and inferotemporal cortices process different types of visual information. It has been suggested that visual information reaching these two subdivisions follows separate pathways from the striate cortex through the prestriate cortex. We examined directly this possibility by placing injections of the retrograde fluorescent tracers, fast blue and diamidino yellow, in inferior parietal and inferotemporal cortex and examining the spatial pattern of cortical areas containing labeled cells in two-dimensional reconstructions of the cortex.

The results of injections in inferotemporal cortex show that TEO receives afferents from areas V2, ventral V3, V3A, central V4, V4t, and DPL in prestriate cortex and from areas IPa, PGa, and FST in the superior temporal sulcus (STS). Area TEp receives afferents only from V4 in prestriate cortex and from IPa, PGa, and FST in the anterior STS. Area TEa receives no prestriate input and is innervated by IPa, PGa, FST, and TPO in the anterior STS.

The results of injections in inferior parietal cortex demonstrate that POa receives afferents from dorsal V3, V3A, peripheral V4, DPL, and PO in prestriate cortex, from MST and *VIP and from IPa, PGa, TPO, and FST in anterior STS. Area PGc (corresponding to 7a) is innervated by PO, MST, and by TPO in the anterior STS.

Examination of the two-dimensional reconstructions of the pattern of labeling after combined injections of fast blue and diamidino yellow in areas POa and TEO revealed that these areas are principally innervated by different prestriate areas. Only a small region, centered on area V3A and extending into V4 and DPL, contained cells labeled by either injection as well as a small number of double-labeled cells. In contrast, areas POa and TEO receive afferents from extensive common regions in the anterior STS corresponding to areas IPa, PGa, and FST.

These results directly demonstrate that visual information from the striate cortex reaches inferior parietal and inferotemporal cortices through largely separate prestriate cortical pathways. On the other hand, both parietal and inferotemporal cortices receive common inputs from extensive regions in the anterior STS which may play a role in linking the processing occurring in these two cortical subdivisions of the visual system.

When negotiating a narrow gap, honeybees tend to fly through the middle of the gap, balancing the distances to the boundary on either side. To investigate the basis of this “centering response,” bees were trained to fly through a tunnel on their way to a feeding site and back, while their flight trajectories were filmed from above. The wall on either side carried a visual pattern. When the patterns were stationary vertical gratings, bees tended to fly through the middle of the tunnel, i.e. along its longitudinal axis. However, when one of the gratings was in motion, bees flying in the same direction as the moving grating tended to fly closer to while bees flying in the opposite direction tended to fly closer to the stationary grating. This demonstrates, directly and unequivocally, that flying bees estimate the distances of surfaces in terms of the apparent motion of their images. A series of further experiments revealed that the distance to the gratings is gauged in terms of their apparent angular speeds, and that the visual system of the bee is capable of measuring angular speed largely independently of the spatial period, intensity profile, or contrast of the grating. Thus, the motion-sensitive mechanisms mediating range perception appear to be qualitatively different from those that mediate the well-known optomotor response in insects, or those involved in motion detection and ocular tracking in man.

Calbindin-D28K and calretinin are homologous calcium-binding proteins localized in many neurons of the central nervous system. We have compared polyclonal antibodies against calbindin and calretinin and have shown by western blots using purified calbindin and calretinin from rat that (1) anti-calretinin does not recognize calbindin and (2) anti-calbindin presents some cross-reactivity with calretinin.

In this report, we have compared by immunohistochemistry the localization of both calcium-binding proteins in the retina of monkey, pig, sheep, rat, cat, pigeon, and salamander. These results are compared with previous data for chick. There are many differences between species and not within species, but some aspects of the distribution are conserved. All species, except rat and monkey, have some cones which contain calbindin only. Most species also have some bipolar cells containing calbindin only. Calretinin is rarely seen in photoreċeptors or bipolar cells. All species have horizontal cells which contain calretinin or calbindin or both. All species have amacrine cells and ganglion cells containing one or other protein.

In the cat ganglion cell layer, the calretinin antisera define a new, asymmetric, type of cell.

The electroretinogram (ERG) of the dark-adapted cat eye in response to brief ganzfeld flashes of a wide range of intensities was recorded after intravitreal injection of n-methyl dl aspartate (NMdlA, cumulative intravitreal concentration of 1.3–3.9 mM) to suppress inner-retinal components, and after intravitreal dl or L-2-amino-4-phosphonobutyric acid (dl-APB, 1–3 mM; l-APB, 1.2 mM) and 6-cyano-7-nitroquinoxaline-2, 3 dione (CNQX, 40–60 µM), to suppress all post-receptoral neuronal responses. Rod PII, the ERG component arising from rod bipolar cells, was derived by subtracting records obtained after APB and CNQX from post-NMDLA records. When we measured the derived response at fixed times after the stimulus, we found that PII initially increased in proportion to stimulus intensity without any sign of a threshold. The leading edge of PII at early times after the stimulus, when the response was still small, was well described by V(t) = kI(t −td)5 where k is a constant, I is the intensity of the stimulus, and td is a brief delay of about 3 ms. Correspondingly, the time for the response to rise to an arbitrary small criterion voltage Vcrit was adequately fitted by tcrit = td + (Vcrit/kI)1/5. The time course of the leading edge of the PII response can be interpreted to indicate that the mechanism generating PII introduces three stages of temporal integration in addition to the three stages that are provided by the mechanism of the rod photoreceptors. This finding is consistent with the operation within the rod bipolar cell of a G-protein cascade similar to that in the rods.

The present work quantifies aspects of photoreceptor structure related to mitochondria, inner segment dimensions, and optical properties, as a basis for furthering our understanding of rod and cone function. Electron-microscopic analyses were performed on the retina of one stumptail macaque (Macaca arctoides) to obtain stereological measurements of ellipsoid mitochondrial density, and sizes and shapes of outer and inner segments. In addition, the distribution of mitochondria and the optical properties of human foveal cones were examined with electron microscopy and Nomarski differential interference contrast (NDIC) imaging. Mitochondria comprised 74–85% of cone ellipsoids and 54–66% of rod ellipsoids in macaque. Ellipsoid volume increased with eccentricity by 2.4-fold for rods and more than 6-fold for cones over eccentricities to 12.75 mm, while the volume of the outer segment supported by the ellipsoid was essentially constant for both rods and cones. Per unit volume of outer segment, cones contained ten times as much mitochondria as rods. In human fovea, as in the rest of the retina, most cone mitochondria were located in the distal inner segment. In the foveal center, however, there are also mitochondria in the myoid, as well as in the outer fiber, proximal to the external limiting membrane (ELM). Analyses of the optical aperture of human foveal cones, the point at which their refractive index clearly differs from the extrareceptoral space, showed that it correlated well with the location of mitochondria, except in the foveal center, where the aperture appeared proximal to the ELM. While mitochondria have an important metabolic function, we suggest that the striking differences between rods and cones in mitochondrial content are unlikely to be determined by metabolic demand alone. The numerous cone mitochondria may enhance the waveguide properties of cones, particularly in the periphery.

Historically, the principal function of vision has been to provide the information needed to support action. Visually mediated actions rely on three systems: the gaze system responsible for locating and fixating task-relevant objects, the motor system of the limbs to carry out the task, and the visual system to supply information to the other two. All three systems are under the control of a fourth system, the schema system, which specifies the current task and plans the overall sequence of actions. These four systems have separate but interconnected cortical representations. The way these systems interact in time and space is discussed here in relation to two studies of the gaze changes and manipulations made during two ordinary food preparation tasks. The main conclusions are that complex action sequences consist of a succession of individual object-related actions, each of which typically involve a turn toward the object (if needed), followed by fixation and finally manipulation monitored by vision. Gaze often moves on to the next object just before manipulation is complete. Task-irrelevant objects are hardly ever fixated, implying that the control of fixation comes principally from top-down instructions from the schema system, not bottom-up salience. Single fixations have identifiable functions (locating, directing, guiding, and checking) related to the action to be taken. Several variants of the basic object-related action scheme are discussed, including single-action events in ball sports involving only one anticipatory gaze shift, continuous production loops in text and music reading, and storage–action alternation in copying tasks such as portrait sketching.