docs/data_dictionary.md

PolyCAT Data Dictionary

This document describes every field in the shared data files.

---

data/metadata/participants.csv

One row per recorded participant.

Field	Type	Units	Description
participant_id	string	—	Participant identifier (e.g., P01)
age	int	years	Age at time of recording
gender	string	—	Self-reported gender (Male / Female)
dominant_eye	string	—	Dominant eye (left / right)
hardware_sampling_rate_hz	int	Hz	EyeLink hardware sampling rate (1000 Hz for all participants). Effective per-eye rate depends on recording mode: 1000 Hz for monocular sessions, 500 Hz for binocular sessions. See recording_mode in sessions.csv for per-session details.
has_part_a	bool	—	Whether Part A data exists
has_part_b	bool	—	Whether Part B data exists
n_trials_total	int	—	Total trials across both parts
n_image_trials	int	—	Number of image (non-empty) trials
n_empty_trials	int	—	Number of empty-aperture trials
included_in_release	bool	—	True if participant is included in the released dataset
exclusion_reason	string	—	Reason for exclusion (empty if included)

---

data/metadata/sessions.csv

One row per session folder in the raw data.

Field	Type	Units	Description
participant_id	string	—	Participant identifier
part	string	—	Experiment part (A or B)
session_id	string	—	Session folder name (e.g., session_20260118_100426)
session_label	string	—	primary, split_primary, or extra
session_timestamp	ISO 8601	—	Parsed from session folder name
n_trials	int	—	Number of data rows in that session's trials.csv
included_in_release	bool	—	True if session contributes to the released dataset
exclusion_reason	string	—	Reason for exclusion (empty if included)
recording_mode	string	—	Eye recording mode: binocular (both eyes, 500 Hz effective per eye), monocular_R (right eye only, 1000 Hz), or no_gaze_file (excluded sessions)

---

data/metadata/trials.csv

One row per trial across all included participants and parts. This is the source-of-truth trial table.

Field	Type	Units	Description
participant_id	string	—	Participant identifier
part	string	—	A or B
session_id	string	—	Session folder this trial came from
trial_uid	string	—	Unique trial ID: {pid}_{part}_b{block:02d}_t{trial:02d}
mini_block	int	—	Block number (1–9)
trial_in_block	int	—	Trial number within block (1–39)
trial_type	string	—	image or empty
image_id	string	—	CAT2000 image identifier (e.g., A_Random_169)
image_path	string	—	Relative path to source image
category	string	—	CAT2000 category (Fractal, Object, OutdoorNatural, Random, Satelite, Sketch)
polygon_id	string	—	Polygon identifier (e.g., polygon_25)
polygon_case	string	—	Polygon group (reference, convexity_varied, or irregular)
polygon_json_path	string	—	Relative path to polygon JSON definition
aperture_scale_factor	int	—	Polygon scaling factor applied during experiment
cue_pos_id	string	—	Fixation cue grid position (e.g., grid_11)
cue_x_px, cue_y_px	float	pixels	Cue position in screen coordinates
cue_x_deg, cue_y_deg	float	degrees	Cue position in degrees from screen center
stimulus_duration_s	float	seconds	Stimulus presentation time (4.0)
iti_s	float	seconds	Inter-trial interval (0.5)
center_screen_x_px, center_screen_y_px	float	pixels	Screen center (1920, 1080)
center_screen_x_deg, center_screen_y_deg	float	degrees	Screen center in degrees (0, 0)
fixation_achieved	bool	—	Whether drift-check fixation was achieved
fixation_attempts	int	—	Number of drift-check attempts
fixation_total_time_s	float	seconds	Total time spent on drift check
ts_trial_start ... ts_trial_end	float	seconds	Key event timestamps (experiment clock)
aborted	bool	—	Whether the trial was aborted

---

data/fixations/fixations_all.csv

One row per fixation event. Fixations detected by EyeLink's built-in algorithm (velocity threshold 30 deg/s, acceleration 8000 deg/s²).

Field	Type	Units	Description
participant_id	string	—	Participant identifier
part	string	—	A or B
session_id	string	—	Session folder
trial_uid	string	—	Unique trial ID
fixation_index	int	—	Fixation order within trial and eye (1-indexed)
eye	string	—	L (left) or R (right)
image_id	string	—	CAT2000 image identifier
category	string	—	CAT2000 category
polygon_id	string	—	Polygon identifier
start_time_ms	float	ms	Fixation start relative to stimulus onset
end_time_ms	float	ms	Fixation end relative to stimulus onset
duration_ms	float	ms	Fixation duration
x_px, y_px	float	pixels	Average fixation position (screen coordinates, top-left origin)
x_deg, y_deg	float	degrees	Fixation position relative to screen center
pupil_size	float	area	Mean pupil size (EyeLink arbitrary units)

---

data/gaze/*.tsv

Sample-level gaze data. One TSV file per session. Tab-separated.

Field	Type	Units	Description
participant_id	string	—	Participant identifier
part	string	—	A or B
session_id	string	—	Session folder
trial_uid_raw	string	—	Raw trial UID from EDF (e.g., B_MB01_T001)
timestamp_rel_ms	float	ms	Time relative to stimulus onset
eye	string	—	L or R
x_px, y_px	float	pixels	Gaze position (screen coordinates)
pupil_size	float	area	Pupil size (EyeLink units)
validity	int	—	0 = valid gaze, 1 = missing/out-of-range

---

data/metadata/quality_metrics.csv

One row per participant. Aggregated quality statistics.

Field	Type	Units	Description
participant_id	string	—	Participant identifier
total_trials	int	—	Total trials for this participant
valid_trials	int	—	Trials with tracking ratio >= 0.8
excluded_trials	int	—	Trials below tracking threshold
tracking_ratio_mean	float	—	Mean tracking ratio (0–1)
tracking_ratio_std	float	—	Std of per-trial tracking ratio
calibration_error_deg_mean	float	degrees	Mean validation error (if available)
fixations_per_trial_mean	float	—	Mean fixation count per trial
fixations_per_trial_std	float	—	Std of fixation count per trial
fixation_duration_mean_ms	float	ms	Mean fixation duration
fixation_duration_std_ms	float	ms	Std of fixation duration
data_loss_percent	float	%	Percentage of trials with no fixations

---

data/manifests/polygon_geometry.csv

One row per polygon shape (27 total). Canonical center coordinates.

Field	Type	Units	Description
polygon_id	string	—	Polygon identifier
polygon_case	string	—	Polygon condition type
json_filename	string	—	Filename of polygon JSON definition

---

data/manifests/stimulus_manifest.csv

One row per trial slot (702 total: 351 per part). Defines the fixed stimulus-polygon-cue assignment shared by all participants.

Field	Type	Units	Description
aperture_scale_factor	int	—	Polygon scaling factor
category	string	—	CAT2000 category
cue_pos_id	string	—	Fixation cue grid position
cue_x_px, cue_y_px	float	pixels	Cue position
image_id	string	—	Image identifier
image_path	string	—	Path to image file
mini_block	int	—	Block number (1–9)
part	string	—	A or B
polygon_case, polygon_id	string	—	Polygon condition and ID
polygon_json_path	string	—	Path to polygon JSON
stimulus_duration_s	float	seconds	Presentation duration
trial_in_block	int	—	Trial position within block
trial_type	string	—	image or empty
trial_uid	string	—	Trial identifier

---

Coordinate Systems

Screen coordinates: Origin at top-left corner of the 3840x2160 display. X increases rightward, Y increases downward.

Degree coordinates: Origin at screen center (1920, 1080 px). Computed using the display geometry:

• 27" diagonal, 3840x2160 resolution, 70 cm viewing distance
• Pixels-per-degree: ~78.5 (computed from ppd = distance_cm * tan(1 deg) / cm_per_px; see docs/acquisition_protocol.md for full derivation)

Time reference:

• timestamp_rel_ms and start_time_ms / end_time_ms: relative to stimulus onset (ms)
• ts_* fields in trials.csv: experiment clock (seconds from session start)

---

Known Data Issues

Issue	Scope	Details
P05 missing trials	Part B only	P05 completed only 317 of 351 Part B trials (34 missing from block 9) due to a session interruption. Total: 668 trials instead of 702.
Aborted trials	48 trials across 9 participants	Trials that were aborted before stimulus presentation. Marked with aborted=True; all timestamp and stimulus fields are NaN. Participants affected: P05 (7), P07 (1), P10 (2), P13 (9), P18 (9), P20 (2), P24 (13), P25 (1), P28 (4).