Abstract
We are interested in the ability of human observers to detect dynamic length changes of computer generated objects and what role attention may play in this task. In a series of experiments we have subjects to detect changes in the length of a two link articulation approximating a human arm moving in the the frontoparallel plane: the upper or “shoulder” end of the articulation remained stationary while the lower or “hand” end moved around a circular path. For lines on paper, the Weber fraction for length differences is 2.9%, and changes in length of the moving arm produced a similar sensitivity of 2.8%. However, changes in only one of the arm segments, rather than the entire arm, produced different sensitivities, 2.9% for the upper segment, and 5.3% for the lower segment (Wilcoxon signed-rank test, p<0.0001). We also asked our subjects to count the number direction changes of the “hand” and judge if the arm increased or decreased in length over a duration of 140–800ms. Again, we found a difference in the sensitivity for changes of the segments: 10.7% for the upper segment and 14.5% for the lower segment (Wilcoxon signed-rank test, p<0.03) while changes in the length of the entire arm produced a Weber fraction of 10.5%. We also found differences in sensitivities for very slowly growing, 6.9%, or slowly shrinking, 19.7%, the arm (Wilcoxon signed-rank test, p<0.0005). Asking subjects to count the number of times a very low contrast distractor changed direction further decreased their sensitivity for slowly growing arms, 15.2%, but not for slowly shrinking arms, 21.3%. These results suggest an intriguing upper bound on the “insensitivity” of the human visual system. While Weber fractions are often presented as averages across subjects, none of the subjects in our experiments have produced a Weber fractions above 30%. Therefore we hypothesize that line lengths are encoded when attention is absent using a coarse resolution that is refined when attention is available.