Fig 1.
Left: an example stimulus image, with randomly distributed red occluding pixels occupying 20% of the image area. Middle: a different character that is occluded by red pixels occupying 60% of the image area. Right: a mask used in the same-different matching task.
Fig 2.
Proportion correct in the same-different matching for the zero, less, and more experienced participants, and for inverted and upright characters.
The error bars represent standard error of the mean.
Fig 3.
Scatter plots of upright (x-axis) vs. inverted (y-axis) same-different accuracies in proportion correct (left panel) and hit rates (right panel).
Each panel shows the 25 occlusion conditions, each of which was averaged from the 105 readers.
Fig 4.
Top: scatter plot for the 105 readers of the same-different matching accuracies with upright characters and the recognition accuracies in the subsequent character recognition task. There was little correlation between the two measures. Bottom: scatter plot of the 105 readers’ years of learning Chinese and accuracies in the character recognition task. The correlation (r = 0.85) was statistically significant.
Fig 5.
Scatter plots of proportion correct accuracies in the same-different matching task for upright and inverted characters for the 51 non-readers (top), and for the 105 readers (bottoms).
The two correlations were statistically significant and comparable to each other.
Fig 6.
Same-different matching accuracies with upright and inverted characters, from three native Chinese speakers who scored perfectly in the character recognition test.
These three native speakers’ same-different matching accuracies were nevertheless not perfect.
Fig 7.
Scatter plot from the 20 native speakers of their same-different matching accuracies with upright and inverted characters.
The correlation coefficient was 0.84.
Fig 8.
Proportion correct accuracies for the 21 native Chinese speakers, for the upright and inverted characters in the same-different matching task.
All 21 participants except No.1 and No.16 showed a higher accuracy for the upright than for the inverted characters. The average difference was statistically significant, but was only 0.01 in proportion correct.
Fig 9.
Proportion correct in the same-different task with upright Chinese characters by 120 non-readers and 29 readers of Chinese.
The accuracies of the two groups were identical. Error bars are standard errors of the mean.