Different authors offer subtly different characterizations of naïve realism. We disentangle the main ones and argue that illusions provide the best proving ground for naïve realism and its main rival, representationalism. According to naïve realism, illusions never involve perceptual error. We assess two leading attempts to explain apparent perceptual error away, from William Fish and Bill Brewer, and conclude that they fail. Another leading attempt is assessed in a companion paper, which also sketches an alternative representational account.