Intuitively, it seems plausible that in real-world scenes, attention is disproportionately attracted by texts. The present study tested this hypothesis and examined some of the underlying factors. Texts in real-world scenes were compared with paired control regions of similar size, eccentricity, and low-level visual saliency. The greater fixation probability and shorter minimum fixation distance of texts showed their higher attractiveness. These results might be caused by the prominent locations or special visual features of text. In another experiment, texts were removed from the scenes, and the results indicated that the locations that used to contain texts did draw more attention than controls. Finally, texts were placed in unexpected positions in front of homogeneous and inhomogeneous backgrounds. These unconstrained texts were found more attractive than controls, with background noise reducing this difference, which indicates that the attraction by specific visual features of text was superior to typical saliency.