"The visual nervous system transforms the input from the visual scene into electrical signals. The electrical signals are in turn read out by other areas of the brain to form perceptual decisions. The relationship between neuronal responses evoked by sensory stimuli and their perceptual correlates is an important research question in modern computational and systems neuroscience. Previous studies of this relationship focused on correlating the response of single neurons to stimuli in their receptive field (RF), defined as a region of the visual space where selective stimuli can evoke a response. However, single neurons are part of a local circuitry that controls their selectivity and modulates their response. The local circuitry is composed of many neurons that encode surrounding regions of the visual space. The neurons in the circuit can interact via excitatory or inhibitory connections. The interactions are crucial for contextual modulation of single neuron response, since any stimulus on the RF of a neuron is part of a larger visual context. Aside from the contextual modulation of neuronal responses, another part of the sensory-to-decision transformation is how the responses of neurons are read out from a population. Correlations in the responses of neurons can have a large impact in the amount of information in a population, and this correlation should be quantified when considering a population readout.Another level of complexity is that downstream areas must use this sensory code flexibly in perceptual decision making. Many neurons and areas can contain the information about the sensory stimulus, and the readout should correspond to the experience of the animal. This thesis investigates these issues in order. In chapter 2, I determine the neural basis for a type of contextual modulation in our motion perception, the worsening of motion perception at large stimulus size. Large stimuli suppress the firing rate of neurons in a form of contextual modulation known as surround suppression. I simultaneously record from multiple neurons in the middle temporal visual area (MT), and I perform a simulation of the recorded responses and the correlation structure. I find surround suppression improves the population sensitivity for small stimuli at the expense of weaker sensitivity for large stimuli. In chapter 3, I examine the underlying circuitry mechanisms for surround suppression. Recent work suggests that the cortex operates in a theoretical network where excitation alone is strong enough to induce instability but inhibition maintains the stability. I pharmacologically manipulate the efficacy of inhibitory processes and find that the neuronal dynamics are consistent with the predictions. I then perform additional experiments to confirm that this network mechanism can be generalized to different stimulus dimensions in MT. In chapter 4, I examine the flexibility of the readout of sensory information. The readout of sensory evidence for visual perception is plastic and depends on recent training experience. I use reversible inactivation and microstimulation to probe the causal relationship between MT neuronal response and perception. I find the causal contribution of MT to visual motion perception depends on training the animals on a specific task of motion integration. This thesis has implications in the broader context of neural coding in health and diseases. Previous work shows that natural aging or disease processes can lead to deficits in our sensory perception, and the reduction of inhibition efficacy has been implicated. Therefore, an understanding of the inhibitory interactions in local cortical circuitry may lead to future treatments and interventions." --