The eye-tracking data is provided in different formats which are explained in this README. The folders are named
accordingly.
One file always contains the data for one reader and one text and is named according to this scheme:
reader[reader_id]_[text_id]_[data_suffix].txt.
"data_suffix" refers to the specific data format.
NOTE: for all the data described below, refer to the CODEBOOK for a full overview of data types, distributions, etc.
The raw data in available in .tsv format. The originally generated .edf files were converted to .asc using the
SR Research edf2asc tool and then parsed to .tsv files containing one sample per line.
Both the .tsv and the .asc files are available. The .ascfiles are in the asc_files folder while the .tsv files
are in the raw_data folder.
NOTE on .asc files: Reader 0 and reader 21 are the same person. Texts 1-11 have been read by the person under ID 0 and text 12 has been read by the person under ID 21. In all the following data files, the reader ID 21 has been changed to 0. This note ONLY applies to the .asc files!
Fixations and saccades were computed from the raw data using the Eyelink Data Viewer software package provided by SR Research with the default parameter settings (SR Research Ltd. 2011). Subsequently, each fixation was mapped to the character index in the text that was fixated annotated with the line index and the character index in the line, the original screen coordinates (in pixels) were discarded. Fixations on the white space between two words were mapped to the closest character. Visual inspection of the data revealed that in certain fixation sequences, vertical calibration error gradually increased over time. This measurement error was semi-automatically corrected by adjusting the fixation-to-character mapping (i.e., re-mapping a fixation to the character in the line above or below the currently mapped character). The files therefore contain the corrected data.
From the fixation data, various reading measures commonly used in reading research were computed. Each word (defined by the surrounding white spaces) was considered one region of interest for all measures except for landing position which was based on the characters within a word. Fixations on the punctuation marks were considered to belong to the preceding word by default and to the following word in case of opening parentheses or opening quotation marks. Definitions of the various measures are provided in the CODEBOOK.md. Note that several of these measures are linearly dependent.
Can be (re)created using this script: additional_scripts/compute_reading_measures.py.
More information on the script is found in the additional scripts' README
This folder contains the reading measures merged with word features and reader information ordered by word order of the original stimulus.
Can be (re)created using this script: additional_scripts/merge_rm_wf.py.
More information on the script is found in the additional scripts' README
This scanpaths folder contains the fixation data ordered temporally by fixation index.
The scanpaths can be (re)created using this script: additional_scripts/generate_scanpaths.py.
More information on the script is found in the additional scripts' README
The scanpath files can be easily merged with the reader information, the reading measures and the word features
using this script additional_scripts/merge_scanpaths_rm_wf.py. The resulting data files will be written to the folder
scanpaths_reader_rm_wf.
More information on the script is found in the additional scripts' README
Original fixation report created from the Data Viewer. The fixations in this report have been manually corrected.
All fixation files contained in this folder are corrected: eyetracking_data/fixations/.
SR Research Ltd. (2011). EyeLink Data Viewer user’s manual [Computer software manual]. Mississauga, Canada. (document version 1.11.1)