Table 1.
Text statistics for the basic text.
Fig 1.
Upper left: A screen from the stimuli. The text was split in 5 screens and had 181 words. Upper right: A participant is seated in front of the monitor. With the support of a chin rest she is reading the text displayed. Bottom left: 25-point calibration screen. Bottom right: 5-point validation screen.
Fig 2.
A reading path from a typical (normal) reader (left) and from a reader with dyslexia (right). The blue circles are the fixations and the orange lines are the saccadic movements. The bigger the circle, the longer the fixation. Clearly the reader with dyslexia exhibits longer fixation duration, shorter saccadic movements, regressive movements and longer reading time than the typical reader.
Table 2.
Parameter τ for the basic text.
Fig 3.
Left: The ROC curve for backward refixations of less than 100 pixels (4 characters). Right: On the x-axis we have the threshold used to count refixations. This number of refixations is used to derive the ROC curve of the parameter, which is used as a classifier. On the y-axis we have the area under this ROC curve. As we can see, we get maximum area at a threshold of 100 pixels, which corresponds to 4 characters. The fitting is done with smoothing splines. Each data (blue) point corresponds to a sensitivity-specificity pair associated to a range of values for the threshold.
Table 3.
Saccade length for the basic text.
Table 4.
Refixations for the basic text.
Table 5.
Fixation number for the basic text.
Table 6.
Not fixated, singly fixated and multiply fixated words for the basic text.
Table 7.
Gaze difficulty based on word length for the basic text.
Fig 4.
Left: Circular validation based on the TRS of the participants. A score above 0.5 (dash line) classifies the participant as typical reader, while a score below 0.5 classifies the participant as potentially dyslexic. Right: Principal component analysis is carried out on the normalized ((P-μ)/σ) vector of all 12 useful parameters of RADAR. The first principal component (maximum variance) vector projection is plotted against the last principal component (minimum variance) vector projection. The separability of readers with dyslexia from typical readers is clear.
Table 8.
Text statistics for the easy text.
Table 9.
Parameter τ for the easy text.
Table 10.
Saccade length for the easy text.
Table 11.
Refixations for the easy text.
Table 12.
Fixation number for easy text.
Table 13.
Not fixated, singly fixated and multiply fixated words for the easy text.
Table 14.
Gaze difficulty based on word length for the easy text.
Fig 5.
A reading path from a typical (normal) reader (left) and from a reader with dyslexia (right). Clearly the dyslexic reader exhibits longer fixation duration, shorter saccadic movements, many regressive movements and longer reading time than the typical reader.
Table 15.
Text statistics for the English texts.
Fig 6.
Age dependency on the parameters.
Non-word-specific parameters dependence on age. Error bars represent standard error of the mean. Although the population size for each grade is small, the trends of the parameters are evident: mean fixation duration (τ) drops with age, saccade length (first quartile plotted here in pixels) increases with age, number of refixations decreases with age, and number of fixations decreases with age. Note that the two populations (dyslexic and typical readers) are clearly separable throughout the entire four-year age span.
Fig 7.
How to choose the best possible test using ROC.
Table 16.
Sensitivity, specificity and threshold values for both texts.