The brain is composed of billions of neurons that exchange diverse patterns of electrical activity. Collectively, these activity patterns encode information about the external environment (such as patterns of light intensity, chemical signals, and acoustic pressure waves), and they ultimately support a vast set of behavioral functions (such as the abilities to see in the dark, to learn to ride a bicycle, and to plan a career path). I am interested in understanding how the anatomical and dynamical organization of neuronal populations facilitates these diverse functionalities, and more generally, why a particular organization is advantageous or disadvantageous. Efficient coding—the notion that information processing systems are efficiently tuned to the statistical regularities of their natural inputs—provides a powerful organizing framework for approaching these questions. I will highlight two applications of efficient coding in the visual system, where it is possible to characterize the statistical structure of visual signals and measure how these signals are encoded in populations of neurons. In one application, we use behavioral experiments to measure human visual sensitivity to correlated textures (visual patterns with controlled statistical properties), and we show that this sensitivity is matched to the distribution of multipoint correlations present in natural scenes. In another application, we use multi-electrode arrays to record the activity of retinal neurons responding to natural and artificial movies, and we show preliminary evidence that the collective structure of population activity patterns is matched to the statistical structure of natural movies. Together, these studies support the hypothesis that the visual system is efficiently tuned to signals in the natural visual environment, and they provide insight into how this match emerges at a population level.