The goals of our laboratory's research are to understand how the brain constructs representations of the environment and how these representations are modified by cognitive processes such as attention, expectation, and learning. We address these questions by studying both healthy human participants as well as patients who suffer from diseases that affect perceptual processing. We use a combination of behavioral, neuroimaging, electrophysiological, modeling, and pharmacological techniques. Specifically, we are investigating the neurophysiological and neurochemical substrates of visual attention and perceptual learning, effects of acetylcholine on perception, memory, and neural representations, visual processing in schizophrenia, binocular rivalry, motion perception, functional subdivisions of the lateral geniculate nucleus, representation of visual space in the brain, and perceptual and neural plasticity in patients with retinal degeneration.


Visual attention

When attention is allocated to a particular portion of the visual field, visual perception is enhanced at that location. Our research in this area is focused on understanding the neurophysiological and neurochemical bases of this enhancement of perception by attention.



Functional connectivity of sustained visual spatial attention

Using functional magnetic resonance imaging (fMRI), we have measured how attention changes the functional connectivity among many identified brain areas. This allows us to generate neural “circuit diagrams” of attention signals in the human brain. In collaboration with Mark D’Esposito and David Heeger, we have shown that sustained visual spatial attention results in increased coupling between areas in posterior parietal cortex (IPS1 and IPS2, Silver et al., 2005) and many visual cortical areas. In addition, attention-related activity in the IPS regions leads activity in several visual cortical areas by a few hundred milliseconds, consistent with transmission of top-down spatial attention signals from IPS1 and IPS2 to early visual cortex (Lauritzen et al., 2009).



Effects of attention on evoked responses and fluctuations in endogenous activity

We have found that continuously allocating spatial attention to a visual stimulus increases the reliability of the fMRI response evoked by the stimulus in a large number of early visual, ventral occipital, lateral occipital, and posterior parietal cortical areas (Bressler and Silver, 2010). This enhanced response reliability by spatial attention is due two a combination of two factors: 1) enhanced amplitude of responses evoked by the visual stimulus, and 2) a reduction in the amplitude of slow endogenous fluctuations in fMRI signals at frequencies below 0.1 Hz. Surprisingly, we have found that the attentional enhancement of evoked responses is not correlated with performance on a target detection task. Rather, behavioral performance is highly correlated with the amount of suppression of slow endogenous fluctuations.



Neurochemical and neurophysiological substrates of voluntary and involuntary spatial attention

We are characterizing neurochemical and neurophysiological differences between voluntary and involuntary spatial attention. Voluntary attention refers to allocation of attention to a location that is relevant for current behavioral goals, while involuntary attention is automatically drawn to the appearance of salient stimuli. In collaboration with Bill Prinzmetal, we have studied the effects of the cholinesterase inhibitor and Alzheimer’s medication donepezil. Donepezil inhibits the enzyme that breaks down acetylcholine in the synapse, thereby prolonging the lifetime of this neurotransmitter. This cholinergic enhancement increases the beneficial effects of voluntary attention but does not affect involuntary attention (Rokem et al., 2010). Ongoing research includes characterization of the brain networks associated with voluntary versus involuntary attention and changes in these patterns of connectivity following cholinergic enhancement with donepezil. We are also using event-related potentials (ERPs) and steady-state visual evoked potentials (SSVEPs) to characterize electrophysiological correlates of surround suppression and their modulation by spatial attention and donepezil.



Location-dependence of attentional modulation of responses to visual stimulation

It is well established that directing attention to a visual stimulus enhances the response evoked by that stimulus in many brain areas. Although many aspects of visual processing differ significantly between central and peripheral vision, little is known regarding the neural substrates of the eccentricity-dependence of spatial attention effects. In collaboration with Lynn Robertson's group, we conducted an fMRI study to measure attentional modulation of stimulus-evoked responses in many topographically-organized cortical areas. We found that in early visual, ventral, and lateral occipital cortex, attentional modulation was greater for central compared to peripheral eccentricities, possibly reflecting a role for attention in resolving fine detail of an attended object in central vision. The opposite pattern was observed in cortical area IPS0, where attentional modulation of positive responses was greater in the periphery, perhaps reflecting the importance of detecting behaviorally relevant objects in the periphery for planning of motor responses.



Effects of feature-based attention on neural representations of faces

We have measured fMRI responses to individual faces along a morph continuum and found that individual voxels (volumes of tissue of several cubic millimeters) can exhibit significant selectivity for individual faces. Also, directing attention to just one of a pair of superimposed faces selectively enhanced responses to the superimposed face pair in voxels previously defined as preferring the attended face. We are currently conducting functional connectivity analyses to identify brain regions that interact with the voxels that are tuned for individual faces.



Hemispheric asymmetries in attentional shifting of visual field representations in topographic parietal cortex

We employed fMRI and the population receptive field (pRF) method to characterize the effects of spatial attention on visual field representations. The pRF represents the portion of the visual field that can evoke a visual response in a given voxel. Attending to a stimulus that traversed the visual field increased pRF size in topographically-organized parietal cortex in both hemispheres. In the left hemisphere, attention also shifted pRFs toward the periphery, thereby maintaining their contralateral representation.In contrast, attention did not shift pRF locations in the right hemisphere, so the increased pRF size resulted in more bilateral representations. These results show that spatial attention can induce bilateral representations in right parietal cortex and offer a new approach for investigating visual attention deficits in hemispatial neglect.



Relationships between spatial attention and episodic memory signals in posterior parietal cortex (PPC)

In collaboration with Anthony Wagner’s laboratory, we employed an episodic retrieval task that identified four distinct regions in PPC that tracked different factors associated with retrieval (Hutchinson et al., in press). Some of the same subjects participated in a spatial attention ask that allowed identification of topographically-organized areas in PPC that exhibit maps of spatial attention signals (Silver et al., 2005; Silver and Kastner, 2009). We found that activity in some of these spatial attention maps (IPS5, SPL1) is related to episodic retrieval outcomes, while others did not. These findings help define the functional organization of PPC and clarify the relationships between neural correlates of episodic memory and spatial attention.



Perceptual learning

Perceptual learning refers to long-lasting improvement in the performance of a perceptual discrimination through training that is specific to the stimuli used for training. It is used as a laboratory model to better understand the processes by which people learn new and intricate skills throughout the lifespan.



Cholinergic enhancement of perceptual learning

We have administered the cholinesterase inhibitor donepezil (see Cholinergic pharmacology section) to healthy human subjects while they were undergoing perceptual learning of motion direction discrimination. This cholinergic enhancement increased the magnitude and specificity of perceptual learning (Rokem and Silver, 2010). A follow-up study with the same subjects showed that both the effects of perceptual learning on motion direction discrimination and the beneficial effects of donepezil on perceptual learning persisted for at least several months after the end of training and drug administration. These results suggest that cholinergic enhancement with donepezil could be used to augment perceptual learning procedures that are used to treat patients with amblyopia. Amblyopia refers to visual impairment in adults that is caused by abnormal visual experience in childhood. In collaboration with Dennis Levi’s laboratory, we are testing whether administration of donepezil during perceptual learning treatment can facilitate the recovery of vision in patients with amblyopia. We are also conducting fMRI/pharmacology studies in healthy human subjects to better understand the neural substrates of perceptual learning and their modulation by acetylcholine.



Acetylcholine, perceptual learning, and sleep

Our finding that training under donepezil increases perceptual learning may reflect a higher fidelity of neural representations of the stimuli used for training (Silver et al. 2008; see also Cholinergic pharmacology section). However, acetycholine also plays critical roles in consolidation of memories during sleep. Donepezil remains in the body for several days after administration, so our results to date are consistent with both encoding- and consolidation-based interpretations. In collaboration with Sara Mednick’s group, we are employing naps and cholinergic pharmacology to directly test possible contributions of acetylcholine to encoding, consolidation, and retrieval of perceptual learning.
Our work with Sara Mednick’s lab also resulted in the discovery of sex differences in the role of sleep in perceptual learning of motion direction discrimination. Men who napped after a single session of training exhibited more learning than men who did not nap, and the perceptual learning was specific to the trained direction of motion in both groups of men. In contrast, women in both nap and no-nap groups showed significant learning of both trained and untrained directions of motion.



Cholinergic pharmacology

We use the cholinesterase inhibitor and Alzheimer’s medication donepezil (trade name Aricept) to enhance the actions of the naturally-occurring neurotransmitter acetylcholine. Blocking cholinesterase activity results in a prolonged lifetime and effectiveness of acetylcholine after it is released by presynaptic terminals and is therefore a useful method for studying the role of this neurotransmitter in cognitive and perceptual processes.



Effects of acetylcholine on perceptual measures of spatial integration

We previously found that donepezil reduces the spatial spread of fMRI responses to visual stimulation in early visual cortex (Silver et al. 2008). This is consistent with research from non-human animal models showing that acetylcholine reduces receptive field size in these cortical areas. Although these data suggest that acetylcholine improves the spatial resolution of stimulus representations in visual cortex, the perceptual consequences of this were not known. We tested the effects of donepezil on surround suppression and crowding, two phenomena that involve spatial interactions between visual field locations. We found that cholinergic enhancement reduces orientation-specific surround suppression of gratings but had no effect on visual crowding of letters (Kosovicheva et al., 2012).



Effects of acetylcholine on information content of stimulus representations

We have developed an fMRI decoding procedure for measuring response reliability across repeated presentations of a movie containing natural scenes. Greater response reliability (i.e., decreased response variance across trials) in a given brain area indicates a more accurate representation of the physical features of the stimulus. Our preliminary results show that cholinergic enhancement with donepezil increases fMRI decoding performance (information content of stimulus representations) in early visual cortex.



Effects of acetylcholine on visual short-term memory

Previous models by Hasselmo and colleagues and our own work on information content of stimulus representations suggest that acetylcholine can increase the fidelity with which stimuli are encoded by the brain. We tested the effects of donepezil on performance of a visual short-term memory task. A set of colored squares was briefly presented, followed by an approximately one second delay period. The subjects were then presented with a second set of colored squares and asked whether any of them had changed compared to the first set. We found that donepezil improved performance on this task, but only for very brief presentations of the first stimulus set. That is, cholinergic enhancement boosted visual short-term memory only under conditions where it was most difficult to encode the stimuli to be remembered, consistent with a role for acetylcholine in improving the accuracy of stimulus representations in the brain.



Visual processing in schizophrenia

In collaboration with Jong Yoon, Richard Maddock, and Cameron Carter at the University of California, Davis, we are studying visual processing in schizophrenia. Patients with schizophrenia have diminished surround suppression, a specific form of contextual modulation of visual perception in which the presence of a surround stimulus decreases perceived contrast of the central surrounded region. Reduced surround suppression allows patients with schizophrenia to perform visual perception tasks in the presence of a surround as well as or even better than healthy controls (Yoon et al., 2009). We have found that the magnitude of surround suppression correlates with the concentration of the inhibitory neurotransmitter GABA in early visual cortex, as measured with magnetic resonance spectroscopy (Yoon et al., 2010). In addition, patients with schizophrenia have reduced levels of GABA in visual cortex relative to control subjects. We have also found that patients with schizophrenia show broader tuning of a perceptual measure of tuning for stimulus orientation, a type of selectivity that has been associated with GABAergic transmission in early visual cortex (Rokem et al., 2011). We are currently conducting fMRI studies to determine the specific cortical circuits that are responsible for reduced surround suppression in schizophrenia.



Binocular rivalry

Binocular rivalry is a phenomenon that occurs when two incompatible images are presented to the two eyes. Even though the visual stimuli remain constant, visual perception alternates between the two monocular stimuli. Binocular rivalry is therefore extremely useful for understanding mechanisms underlying the selection of visual inputs for perception. We are studying the mechanisms of perceptual selection in binocular rivalry as well as the effects of predictive context and volitional control.



Roles of magnocellular and parvocellular systems in perceptual selection

The magnocellular and parvocellular streams are two early visual pathways that are specialized for the processing of motion and form, respectively. We have documented differential contributions of these two processing streams to perceptual alternations during binocular rivalry (Denison and Silver, 2012). We studied a phenomenon called interocular switch rivalry in which rival images that are periodically exchanged between the two eyes can generate different types of perceptual alternation. By varying the spatial, temporal, and luminance properties of the rivalrous stimuli, we could bias processing towards either the magnocellular or parvocellular stream. These stimulus manipulations produced reliable changes in the type of perceptual alternation experienced by subjects during interocular switch rivalry, establishing a framework for investigating the differential contributions of the magnocellular and parvocellular systems to visual perception.



Predictive context in binocular rivalry

In this work, we present stimuli that engender expectations of forthcoming stimuli by the subject and then measure how these expectations influence which of the rivaling images is selected by the brain for conscious perception. In one study, presentation of a moving stimulus to both eyes creates the expectation that the stimuli will continue moving in the same direction. After exposure to the moving stimulus, the subject is presented with a pair of rival images, one consistent with the expectation and one that is inconsistent. We have found that the expected image is more likely to be perceptually selected (Denison et al., 2011).



Volitional control of perception during binocular rivalry

We are characterizing the conditions for which subjects have volitional control over perceptual selection in binocular rivalry. We are employing a variety of binocular rivalry stimuli that are processed at different levels of the visual hierarchy in the brain in order to characterize the factors that influence subjects’ ability to consciously control their percepts during binocular rivalry.



Hemispheric specialization in perceptual selection in binocular rivalry

The double filtering by frequency (DFF) theory developed by Lynn Robertson and Rich Ivry proposes that the two cerebral hemispheres process different spatial frequency components of the visual environment. Specifically, the left hemisphere is specialized for higher spatial frequencies, with the right hemisphere preferentially processing lower spatial frequencies. We have tested whether this theory also applies to perceptual selection. We bias processing of rivalous gratings of different spatial frequencies to one of the hemispheres by presenting them in either the left or right visual hemifield. The likelihood of perceptual selection of a particular spatial frequency depends on whether the rivalrous stimuli are primarily processed by the left or right hemisphere, in accord with the DFF theory.



Computational modeling of motion perception

Humans perceive visual stimuli moving along the cardinal axes (up/down, left/right) better than stimuli moving along oblique (diagonal) directions, a phenomenon known as the oblique effect. We measured the directional tuning of adaptation to moving visual stimuli and found a corresponding oblique effect: the tuning width of adaptation was smaller for cardinal adapting stimuli relative to oblique adapters. We constructed a computational model of encoding of motion stimuli by cortical areas V1 and MT and decoding of stimulus information from the cells in MT (Rokem et al., 2009). This model accounts for a number of properties of directional anisotropies in motion perception and suggests that oblique effects could arise from anisotropies in stimulus encoding in the visual system combined with a decoding mechanism that employs a statistically optimal strategy to read out this information.



Functional subdivisions of the lateral geniculate nucleus (LGN)

Parallel processing in the visual system is particularly well-defined in the LGN of the thalamus, with the magnocellular subdivision selective for high temporal frequencies and low spatial frequencies and the parvocellular subdivision selective for low temporal frequencies, high spatial frequencies, and color. However, these subdivisions have rarely been studied in humans, due to their small size and location deep in the brain. In collaboration with David Feinberg and Essa Yacoub, we have employed high-resolution fMRI and selective visual stimulation to reliably localize the magnocellular and parvocellular subdivisions of the LGN in the human brain. This will facilitate investigations into the role of these specialized early visual pathways in human visual perception, attention, and cognition.



Representation of visual space

Subjects’ estimation of the eccentricity of stationary targets in the peripheral field often exhibits systematic biases. In collaboration with Lynn Robertson's group, we characterized the influence of visual boundaries on these localization biases and discovered fundamental differences between intrinsic boundaries (the edges of the visual field) and external boundaries (the nose, brow, or experimentally imposed boundaries) (Fortenbaugh et al., 2012). Specifically, we observed peripheral localization biases when the visual field was not bounded by external borders but foveal biases when external boundaries were present.



Space perception and its neural substrates in patients with retinal degeneration

Retinal degeneration is the leading cause of visual disability in adults in industrialized societies. In collaboration with Lynn Robertson and Wayne Verdon, we are using psychophysical methods to measure distortions in space perception in patients with retinal degeneration. In parallel, we are characterizing the representations of the visual field in a number of cortical areas with fMRI and correlating distortions in these visual field maps with the perceptual measurements.
In other work, we are studying whether the reduced spatial integration by donepezil could be used to alleviate some of the perceptual symptoms of macular degeneration. Patients with macular degeneration have impaired vision in the central part of their visual field and must rely on peripheral vision that has poor resolution of fine spatial detail. If donepezil enhances the resolution of peripheral vision in these patients, this should result in improved face recognition, reading, and other important visual functions affected by this disease.