The code in this repository is a reimplementation of Nyström, M. & Holmqvist, K. (2010), "An adaptive algorithm for fixation, saccade, and glissade detection in eye-tracking data". Behavior Research Methods 42(1): 188-204. It processes the recorded eye movement data to extract saccades, fixations, and glissades (the latter are now often called post-saccadic oscillations). When using this code, in addition to Nyström & Holmqvist (2010), please cite Niehorster, Siu & Li (2015). See below.
First, the internals of the algorithm have been rewritten extensively with an eye on increasing performance. Furthermore, quite a few additions have been made. This is a non-exhaustive list:
- optionally (and by default) less strict about NaN/missing data during events.
- calculates a lot of information about classified events.
- option to detrend the gaze velocity data by median filtering it and then subtracting the output of this median filter from the gaze velocity trace.
- option to cross-correlate (detrended or original) gaze velocity data with a saccade template for additional noise robustness in saccade classification.
- can detect blinks by means of thresholding change-of-pupil size signal. This enables also detecting partial blinks.
- option to merge nearby saccades/glissades
- option to determine saccade onset by fitting straight line to accelerating flank of velocity profile of saccade, and seeing where this line intersects with vel==0. This was used to get precise saccade onset times in a low-sampling frequency data (see Oliva et al. (2017) below).
- includes post-processing utilities to merge glissades into their associated saccades, remove saccades from the position and velocity traces, and keep only the saccades in the position and velocity traces.
- detailed plotting functionality
When using this code, please cite Niehorster, Siu & Li (2015). If using ETparams.saccade.onsetRefineMethod=2, please additionally cite Oliva, Niehorster, Jarodzka & Holmqvist (2017). Example citation:
Saccades were classified using the Niehorster, Siu & Li (2015) implementation of the Nyström & Holmqvist (2010) algorithm, with default settings. In addition, saccade onsets were determined using the method of Oliva, Niehorster, Jarodzka & Holmqvist (2017).
NB: it is probably good to discuss these methods in a few lines each. It is furthermore important that if you change settings from their default, you note these changes in your article.
References:
Nyström, M. & Holmqvist, K. (2010), "An adaptive algorithm for fixation, saccade, and glissade detection in eye-tracking data". Behavior Research Methods 42(1): 188-204. doi: 10.3758/BRM.42.1.188
Niehorster, D.C., Siu, W.W.F., & Li, L. (2015). Manual tracking enhances smooth pursuit eye movements. Journal of Vision 15(15), 11. doi: 10.1167/15.15.11
Oliva, M., Niehorster, D.C., Jarodzka, H., & Holmqvist, K. (2017). Social Presence Influences Saccadic and Manual Responses. I-Perception 8(1). doi: 10.1177/2041669517692814
Example data from two experiments is provided with this implementation.
- In the folder
NiehorsterSiuLi2015, data from two participants from one condition from Niehorster Siu & Li (2015) (citation above) is provided. This dataset contains pursuit and saccades. RuneventClassificationNiehorsterSiuLi2015.mto run event classification, including creating of data traces containing only pursuit and only saccades. - In the folder
pictureViewing, data from two participants, 10 trials each, is provided, recorded during an unpublished experiment. This dataset contains fixations and saccades. RuneventClassificationPictureViewing.mto run event classification, including implicit classification of fixations in the data.