goxpyriment

Letter Height Superiority Illusion

Replication of:

New, B., Doré-Mazars, K., Cavézian, C., Pallier, C., & Barra, J. (2015). The letter height superiority illusion. Psychonomic Bulletin & Review, 22(4), 1010–1018. https://doi.org/10.3758/s13423-014-0753-8

Participants compare the heights of two briefly presented stimuli and decide which one is taller (or whether they are the same height). The key finding is that letters are perceived as taller than pseudoletters or mirror-image letters of identical physical size.

Experiments

Experiment 1 – Letters (`-exp 1`)

Parameter	Value
Stimuli	9 lowercase letters: a c e m r s v w z
Controls	Mirror letter (horizontal flip) · Pseudoletter (approximation)
Sizes	Small: 0.28° · Tall: 0.30° visual angle
Eccentricity	±2.75° from centre
Stimulus duration	700 ms
Total trials	648 (216 unique × 3 repetitions)
Breaks	Every 108 trials
Training stimuli	u, n, x

Experiment 2 – Words (`-exp 2`)

Parameter	Value
Stimuli	9 uppercase French words: BATEAU BUREAU CAMION CANAL GENOU JARDIN LAPIN PARFUM TUYAU
Controls	Mirror word · Reversed-syllable pseudoword · Nonword (approximation)
Sizes	Small: 0.40° · Tall: 0.44° visual angle
Eccentricity	±0.90° from centre
Stimulus duration	500 ms
Total trials	864 (288 unique × 3 repetitions)
Breaks	Every 144 trials
Training stimuli	RADIO, PAPIER, MAISON

Prerequisites

Go 1.25+
SDL3 development libraries (sudo apt install libsdl3-dev on Ubuntu/Debian)
A display — ideally a 17-inch 1024×768 monitor at 64 cm viewing distance to match the original study. On any other setup the visual angles will be approximate.

Running

# Experiment 1, participant 12, fullscreen
go run main.go -exp 1 -s 12

# Experiment 2, participant 12, fullscreen
go run main.go -exp 2 -s 12

# Windowed 1024×768 (development / testing)
go run main.go -exp 1 -s 0 -d

Flags

Flag	Default	Description
`-exp`	`1`	Experiment number: `1` (letters) or `2` (words)
`-s`	`0`	Participant ID (integer)
`-d`	off	Development mode: windowed 1024×768 instead of fullscreen

Trial structure

Fixation cross  →  Two stimuli  →  Blank screen + response  →  ITI
    200 ms           700/500 ms        (wait for key)           750 ms

Response keys

Key	Meaning
← Left arrow	Left stimulus is taller
→ Right arrow	Right stimulus is taller
↓ Down arrow	Both stimuli are the same height

Training

Before the main experiment the participant completes a practice phase using stimuli not shown in the main session (letters u, n, x for Exp 1; words RADIO, PAPIER, MAISON for Exp 2). Only same-category pairs (letter vs. letter, word vs. word) are used, and feedback (CORRECT / WRONG) is given after each trial. The practice loops until 80 % accuracy is reached.

Output

Data are saved to goxpy_data/ as a .xpd file (CSV with a metadata header). One row per trial:

Column	Description
`trial`	Trial number (1-based)
`anchor`	The letter or word always rendered normally
`anchor_pos`	Position of the anchor: `left` or `right`
`comp_type`	Comparison type: `letter` · `mirror` · `pseudoletter` · `reversed_syllable` · `nonword`
`height_cond`	`same_small` · `same_tall` · `anchor_tall` · `anchor_small`
`response`	Participant’s response: `LEFT` · `RIGHT` · `SAME`
`rt_ms`	Reaction time in milliseconds (stimulus offset → key press)

Notes on stimulus fidelity

Font: Liberation Serif Regular is used as a metric-compatible substitute for Times New Roman (the original font). The font file is embedded in the binary.
Mirror stimuli are produced by horizontally flipping the rendered letter/word texture, which exactly matches the paper’s “vertical mirror symmetry transformation”.
Pseudoletters (Exp 1) and nonwords (Exp 2) are approximated by vertically flipping the source texture. The original study used custom-drawn glyphs carefully matched in height, pixel count, and contiguity. For a faithful replication, obtain the original stimuli from the authors and load them as image files.
Reversed-syllable pseudowords (Exp 2) are the actual text strings from the paper’s appendix (TEAUBA, REAUBU, MIONCA, NALCA, NOUGE, DINJAR, PINLA, FUMPAR, YUTAU).
Visual angles are computed for a 17-inch 1024×768 screen at 64 cm viewing distance (75.3 DPI). On a different screen or distance the physical sizes will differ.

This site is open source. Improve this page.