Glossary¶

The terms the app and these docs use. A measure links to its entry in Computations & methodology, which gives the exact formula, its edge cases, and how far it has been verified.

Eye movements¶

Fixation: A pause of the eyes on one spot. Each row of the fixation table is one; the figure draws it as a marker sized by duration, on one scale shared by every figure.
Saccade: The jump from one fixation to the next, drawn as a line between them. The app classes each one by its role in reading — forward, skip, refixation, return sweep or regression (assign.saccade_class).
Regression: A saccade back to an earlier word (assign.progression).
Return sweep: A saccade down to a later line — normally from the end of one line to the start of the next.
Scanpath: One reading's fixations in order, with the saccades between them.
Raw gaze: The tracker's sample-by-sample gaze positions, from before fixations were detected. An optional third table, drawn under the fixations — or a dataset's only table, drawn as recorded. Scanpath Studio detects no fixations from it.
Run, pass: Consecutive fixations on one word form a run; a word's first run is its first pass, its second run its second pass (assign.runs).

Words and the screen¶

Word box: The rectangle a word occupied on screen — its area of interest (AOI), or interest area (IA) in EyeLink's terms. It is taken from the data exactly as given, never recomputed (geom.word_box_bounds). A fixation counts for the word its data names when the data has a word/IA id (a blank there means no word); without one, for the word whose box contains it, else for no word (assign.fixation_to_word).
Words (interest areas): The table of words and their word boxes, such as EyeLink's Interest Area Report.
Canvas: The recorded screen in pixels, such as 2560 × 1440 — the coordinate system every figure is drawn in.
True to scale: Text and fixations drawn at their recorded on-screen positions, with each word label sized from the word boxes (disp.true_scale).
Recording setup: How the text was shown: the screen resolution (the canvas), the monitor's size and viewing distance, and the text size, each marked measured, estimated or assumed. Size and distance give pixels per degree of visual angle (geom.pixels_per_degree).
Screen: One of several displays a single trial was read over — the pages of a long text, or its comprehension-question screens. Each screen keeps its own coordinate space and is never merged with another (Data format).
Critical span: A marked stretch of the text — in OneStop, the words that answer the trial's question (is_in_aspan) — highlighted in the figure.

Reading measures¶

Scanpath Studio shows these as your interest-area report provides them. The definitions below are how it would compute them (experimental); the values you load mean what your eye-tracking software defines.

First fixation duration (FFD): How long the first fixation on a word lasted, whenever it came (measure.ffd).
First-pass reading time (FPRT): Also called gaze duration: the sum of the fixations in the word's first run, before the eyes first leave it (measure.fprt).
Regression-path duration (RPD): Also called go-past time: from the word's first fixation until the first fixation on a later word, every regression in between included (measure.rpd).
Total fixation duration (TFD): Every fixation on the word, in any pass (measure.tfd).
Single-fixation duration: The first fixation's duration, for a word whose first run was exactly one fixation (measure.single_fix).
Second-pass duration: The sum of the fixations in the word's second run (measure.second_pass).
Skip: A word not fixated before the eyes first moved past it; a word first reached by a regression still counts as skipped (measure.skip).
Initial landing position: Where in the word the first fixation landed, in letters from its start (measure.landing_position).

The app's own terms¶

Trial: One reading: a participant and a trial id together. The same text read by two people is two trials.
Text: The stimulus (text_id) — what two trials of the same text share.
Dataset: One loaded corpus: the bundled demo, a public corpus, or tables you uploaded, each listed under Data Management.
Participant metadata: An optional table with one row per participant, whose columns become trial filters without being copied onto the words or fixations.
Illustration: A label the app puts on any figure that no longer shows the data exactly as recorded — snapped fixations, arced saccades, a subset of the fixations, a replay not at real time (disp.illustration).