Web port was text-on-text matching (target word + 4 word options).
Unity Vocabulary is spoken-word -> picture matching with 4 image
options. Reauthored.

Stimuli: 89 .jpg sprites copied from Assets/Sprites/Vocabulary/ +
30 .ogg audio clips (FR + IT, the 1 practice + 14 test trials that
survive condensation) from Assets/Addressables/VoiceOver/Vocabulary/.
2 tutorial + 14 test items per locale built from Vocabulary.unity
GUID -> filename map (option[0] = correctAnswer in stimuli.json).

RT measurement now starts after audio end + 1s PresentationDelay
(audio-armed click gating, mirrors Unity's SetInteractionTime); was
trial-onset.

Shuffle: re-implemented Unity's 16-config A/B/C/D shuffle tables from
VocabularyStage._shuffleConfigs[A-D] including the
"A doesn't repeat its slot twice" constraint.

Trial flow: 2 aided + 2 unaided trials condensed to single instructions
card + 1 practice item with feedback. 6s soft / 10s hard / 140s block
timers; <1s alarm counter capped at 5; replay button per Unity btn.

CSV: disposition (Unity answer_1;answer_2;...), correct_slot,
chosen_slot, soft_timer_triggered. trial_type enum [practice, test];
analysis-relevant columns blanked on the practice row.

Localisation: FR + IT only (Unity ships en/de blank).

stimuli.json cleanup: removed tut_02 entries and 4+1 orphan images.
0 missing refs, 0 orphans.