Are we simply dividing number of characters in a text with the number of saccades? Should we divide saccade length into 2 categories as regressive saccade length and forward saccade length?
first of all, I personally would not simply divide the number of characters by the number of saccades as that would not give you saccade length, rather a sort of "character/saccade ratio".
Assuming that you want to calculate the saccade length (or saccade amplitude) for a given reading trial of a given participant, here's what you need to know:
(1) As a standard, saccade amplitude is usually expressed in degrees (°) of visual angle (http://en.wikipedia.org/wiki/Visual_angle)
(2) From your eye tracking data, extract saccade events details such as temporal and spatial coordinates, i.e. when (e.g. timestamp in ms) and where (e.g. x and y pixel coordinates) each saccade starts and ends (assuming you have quick access to saccade events through your eye tracker software: most systems are sold along with event-detection software, if this is not the case you would have to extract fixations and saccades from the raw data, you can have a look here in case: https://www.cs.drexel.edu/~salvucci/publications/Salvucci-ETRA00.pdf)
(3) You might want to divide saccades in (a)progressive saccades (i.e. saccades in the direction of reading), (b)line returns (i.e. saccades in the opposite direction, landing to the next line of text), (c)regressive saccades (i.e. saccades in the opposite direction, landing to any previous line of text).
This can be done assuming you have - for each saccade - temporal and spatial coordinates.
(4) For each saccade, calculate the amplitude, for example as euclidean distance between its starting and ending point. Of course, all this should be done separately for each of the three classes identified in step (3).
Depending on which eye tracking system you are using, you might not need to calculate saccade amplitude yourself, as most commercial systems already provide you with this information (e.g. Tobii and SMI systems). What you might need to do is converting from pixels to degrees of visual angle (here you'll find a very clear example: http://osdoc.cogsci.nl/miscellaneous/visual-angle/ ).
For progressive saccades, you should end up with an average value of approximately 2° (for silent reading in "standard" conditions). I also would suggest you to read this comprehensive review on eye movements in reading if you want to know more: http://www.tandfonline.com/doi/abs/10.1080/17470210902816461#.VCp0ahZQSIA
first of all, I personally would not simply divide the number of characters by the number of saccades as that would not give you saccade length, rather a sort of "character/saccade ratio".
Assuming that you want to calculate the saccade length (or saccade amplitude) for a given reading trial of a given participant, here's what you need to know:
(1) As a standard, saccade amplitude is usually expressed in degrees (°) of visual angle (http://en.wikipedia.org/wiki/Visual_angle)
(2) From your eye tracking data, extract saccade events details such as temporal and spatial coordinates, i.e. when (e.g. timestamp in ms) and where (e.g. x and y pixel coordinates) each saccade starts and ends (assuming you have quick access to saccade events through your eye tracker software: most systems are sold along with event-detection software, if this is not the case you would have to extract fixations and saccades from the raw data, you can have a look here in case: https://www.cs.drexel.edu/~salvucci/publications/Salvucci-ETRA00.pdf)
(3) You might want to divide saccades in (a)progressive saccades (i.e. saccades in the direction of reading), (b)line returns (i.e. saccades in the opposite direction, landing to the next line of text), (c)regressive saccades (i.e. saccades in the opposite direction, landing to any previous line of text).
This can be done assuming you have - for each saccade - temporal and spatial coordinates.
(4) For each saccade, calculate the amplitude, for example as euclidean distance between its starting and ending point. Of course, all this should be done separately for each of the three classes identified in step (3).
Depending on which eye tracking system you are using, you might not need to calculate saccade amplitude yourself, as most commercial systems already provide you with this information (e.g. Tobii and SMI systems). What you might need to do is converting from pixels to degrees of visual angle (here you'll find a very clear example: http://osdoc.cogsci.nl/miscellaneous/visual-angle/ ).
For progressive saccades, you should end up with an average value of approximately 2° (for silent reading in "standard" conditions). I also would suggest you to read this comprehensive review on eye movements in reading if you want to know more: http://www.tandfonline.com/doi/abs/10.1080/17470210902816461#.VCp0ahZQSIA
Thanks for your great answer. I calculated pixels for each saccade depending on the screen resolution and participant distance. And by using these pixels I calculated the visual angle for a short line in the text which included 7 progressive saccades. See here: (0,68-8,46-2,91-6,42-5,28-4,89-2,60). The average visual angle is 4,46. This number equals to 136 pixels in my screen. And this 136 pixels equal to 8 characters as average saccade length. Is that procedure correct? Thanks a lot.