Reading Measures

Reading measures are word-level eye-movement metrics calculated from AOI-mapped fixations. The pipeline computes them per trial by first annotating fixations with run and pass information, then deriving a set of standard measures for each word. The implementation lives in metrics.reading.reading_measures.

Fixation Annotation

Every fixation gets tagged with a run ID, first-pass status, regression flags, and neighbouring word indices. This annotation is the foundation for all measures. The pipeline groups fixations by trial, stimulus, and page, then computes:

• run_id — contiguous fixations on the same word form a run. Each time the reader moves to a different word and comes back, a new run starts.

• is_first_pass — a run is first-pass if the word is entered from the left, no higher word index has been fixated before, and the word has not been entered before. All later revisits (regressions-in) are marked as non-first-pass.

• is_reg_in / is_reg_out — a regression occurs when the next fixation lands on a lower (regression-out) or the previous fixation came from a higher (regression-in) word index.

• is_first_fix — the very first fixation on the word in the entire trial.

• prev_word_idx / next_word_idx — word indices of the preceding and following fixation, used for saccade length calculation.

Fixation-based Measures

Abbreviation	Description
FFD	Duration of the first fixation on a word during first-pass reading, 0 if skipped in first-pass
FD	Duration of the first fixation on the word, regardless of pass
FPF	1 if the word was fixated in first-pass, 0 otherwise
FPFC	Number of first-pass fixations on the word
FPRT	Sum of all first-pass fixation durations (0 if skipped)
FRT	Sum of fixations from first entering the word until first leaving it (same as FPRT when the first run was first-pass)
TFC	Total number of fixations on the word
TFT	Sum of all fixations (FPRT + RRT)
RR	1 if the word had re-reading after first-pass, 0 otherwise
RRT	Sum of fixations that are not part of first-pass (TFT - FPRT)
SFD	Duration of the only first-pass fixation; 0 if skipped or multiple first-pass fixations

Transition-based Measures

Abbreviation	Description
LP	Character index of the first fixation on the word (landing position)
RBRT	Sum of fixation durations on the word until a word to its right is fixated (RPD_inc - RPD_exc)
RPD_exc	Sum of regressive fixation durations after a first-pass regression until fixating a word to the right, excluding fixations on the word itself
RPD_inc	Sum of all fixations from first first-pass fixation on the word until fixating a word to its right (includes regressive fixations); 0 if not fixated in first-pass
SL_in	Saccade length leading to the first fixation on the word (in words); positive = progressive, negative = regressive
SL_out	Saccade length of the first saccade leaving the word (in words)
TRC_in	Number of regressive saccades landing on this word
TRC_out	Number of regressive saccades initiated from this word

Output Format

Reading measures are saved as one CSV per trial in reading_measures/<sid>/. The filename follows the pattern {sid}_{trial}_{stimulus}_reading_measures.csv. Each row is a word, with columns for the trial identifier, page number, word index, word text, and all measures above. Words that were never fixated have zeros in all measure columns and are marked as skipped = 1.

AOI Enlargement

AOIs are enlarged vertically before mapping to cover the space between lines. The original AOIs from the stimulus files only cover the exact character bounding box. A fixation slightly above or below the line would miss the word. The enlarge_aois() function in mapping.aoi extends each AOI upward and downward by half the inter-line spacing, so fixations between lines still map to the correct character.