Response bias in very first judgments of the reflectance of grays: Numerical versus linear estimates

Three hundred unsophisticated Os judged the lightness of a gray, reflectance 44%, against a white, reflectance 83%. Half gave a numerical estimate calling the white 100. The remainder marked a 100 mm line labeled 100 at one end and zero at the other. Both distributions of estimates were skewed with a mode around 60. The linear estimates tended to avoid the middle of the line (p<.01).The numerical estimates tended to end in zero or five (p<.001). The two distributions were not quite reliably different, but the medians of subgroups of 10 Os drawn from the two distributions were reliably different (p<.05).