Skip to content

goxpyriment compared with PsychoPy

PsychoPy is the established, most widely used platform for creating and running psychological experiments. This page compares it with goxpyriment feature by feature, so that you can judge whether goxpyriment fits your paradigm — and, where it does not, what you would be giving up.

The comparison is based on the PsychoPy documentation sources (psychopy/psychopy-docs, release branch) and on the current state of this repository.

Summary in one sentence: PsychoPy optimises for accessibility and breadth (a GUI builder, online deployment, support for nearly every laboratory device); goxpyriment optimises for deployability and timing determinism (one static binary, no interpreter, no garbage collection inside the presentation loop).

If you are porting existing PsychoPy code, read the Migration Guide — it gives a concept map and side-by-side code for a representative trial sequence.

The structural difference

PsychoPy is three layers stacked on top of each other:

  1. Builder — a graphical editor in which an experiment is assembled from routines, loops, and ~45 drag-and-drop components. It compiles to Python or to JavaScript.
  2. Coder — the psychopy.* Python library, which the Builder generates calls into and which you can use directly.
  3. PsychoJS / Pavlovia — a JavaScript runtime plus a hosting platform, with integrations for Prolific, MTurk, Qualtrics, and Sona.

goxpyriment is only the middle layer: a Go library. There is no graphical builder and no hosting platform. That single fact explains most of the gaps listed below — and most of the advantages.

Where goxpyriment matches or has an edge

Aspect PsychoPy goxpyriment
Deployment Python environment plus wheels, or Pavlovia hosting A single self-contained binary. SDL3, SDL3_ttf and SDL3_image are embedded as compressed blobs and loaded at startup — nothing to install on the target machine
Presentation loop Python, subject to the GIL and to interpreter overhead; Psychtoolbox/ioHub back-ends are used for precision VSYNC-locked loops with the garbage collector disabled (debug.SetGCPercent(-1)) for the duration of the critical section
Clock psychopy.clock, core.getTime clock package built on SDL's high-resolution performance counter
RSVP and rapid streams Written by hand as a frame-by-frame loop A dedicated high-precision path, stimuli.PresentStreamOfImages, with variants for images, text, tones, mixed streams, and per-frame callbacks
Adaptive procedures StairHandler, QuestHandler staircase.UpDown (Levitt 1971) and staircase.Quest (Watson & Pelli 1983) — the same two families
Unit conversion psychopy.monitors, deg/cm/norm units units package: pixels ↔ degrees ↔ cm through a Monitor struct
Counterbalancing CounterbalanceRoutine (a recent addition) design.AddBWSFactor / GetPermutedBWSFactorCondition — Latin-square between-subject assignment, present from the start
Gamma correction monitors calibration apparatus.Gamma
Data output .csv / .psydat / Excel .csv with a #-prefixed metadata header, written through a buffered output file
System reporting psychopy.info sysinfo package: CPU, GPU, audio, memory, machine — including a WMI path on Windows
Hardware triggers Parallel, serial, LabJack, Cedrus, EGI, BrainProducts, TriggerBox Parallel port (Linux LPT), DLP-IO8, FT232H, Linux GPIO (Raspberry Pi and other SBCs), LabJack T4 over Modbus TCP, NeuroSpin MEG TTL box, generic serial
Networked recording control Egi/iohub, emulator, MQTT/OSC components EGI NetStation event markers over TCP/IP (ECI), plus a "BEL_video" participant-video recorder client
Worked examples A demos folder ~50 complete paradigms in examples/, plus a tests/ suite of programs whose output is analysed to verify timing and hardware behaviour

The core of an experiment — window, text, images, shapes, gratings, dot fields, keyboard and mouse input, sound, trial structure, randomisation, staircases, data files — is covered by both.

Where PsychoPy is ahead

1. Eye tracking

This is the single largest gap. PsychoPy's ioHub provides a vendor-neutral API over EyeLink, Tobii, GazePoint and Pupil Labs, with Builder routines for calibration and validation and a region-of-interest component.

goxpyriment has no eye-tracking support at all — no gaze stream, pupillometry, calibration, or region-of-interest components. (It does ship a triggers.VideoRecorder client for a networked participant-video recorder, but that only records and labels footage of the participant; it is not gaze tracking and returns no eye position.)

2. Breadth of visual stimuli

PsychoPy's visual module documents 42 classes. Notable ones without a goxpyriment equivalent:

  • ElementArrayStim — thousands of elements drawn in a single call. Required for Glass patterns, large-scale multiple-object tracking, and texture-defined stimuli. Probably the most consequential omission after eye tracking.
  • Aperture — restricting drawing to an arbitrary region.
  • RadialStim, NoiseStim, SecondOrder — specialised psychophysical textures.
  • Form, Slider, RatingScale — questionnaire and rating widgets. goxpyriment offers stimuli.ChoiceGrid, stimuli.ThermometerDisplay and stimuli.TextInput, which cover some but not all of this ground.
  • BufferImageStim — snapshotting a rendered screen for fast redisplay.
  • windowWarp — spherical and cylindrical warping for domes and curved screens.
  • The entire 3D/VR branch: rift, SphereStim, ObjMeshStim, PhongMaterial, LightSource, SceneSkybox.

3. Online experiments

PsychoPy's online/ documentation covers Pavlovia synchronisation, participant-recruitment integrations (Prolific, MTurk, Qualtrics, Sona), credit and statistics dashboards, and compatibility checking.

goxpyriment compiles to WebAssembly and runs in the browser (see WASM.md), so the runtime half of this exists. What does not exist is the surrounding service: hosting, recruitment, and server-side data collection are left to you.

4. Video

PsychoPy plays whatever ffmpeg can decode. goxpyriment plays MPEG-1 directly (pure Go, with MP2 audio) and requires conversion to .gv for timing-critical clips — cmd/gv-convert does this, needing no external tools for MPEG-1 input and ffmpeg for anything else. See UserManual §15.

The .gv requirement is a deliberate trade: independent per-frame blocks remove the decode-cost asymmetry between I- and B-frames, which is what makes H.264 onset timing jitter periodically. But it is still friction, the files are roughly an order of magnitude larger, and it rules out playing arbitrary participant-supplied media at measured onsets.

5. Ecosystem and tooling

  • A plugin architecture for third-party components.
  • The Session API for running several experiments under one process.
  • Alerts and linting that catch common design errors before a run.
  • A logging framework with configurable levels.
  • A substantial teaching corpus and community.

6. Camera and speech

PsychoPy has a CameraComponent, Google Speech integration, and audio/visual validator routines. goxpyriment provides apparatus.Microphone, apparatus.VoiceKey and WAV writing, but no camera capture and no speech recognition.

7. The Builder itself

For readers who do not program, the Builder is not a convenience — it is the only way in. goxpyriment requires writing Go.

Choosing between them

Reach for PsychoPy when you need eye tracking; when the paradigm depends on large element arrays, apertures, or 3D/VR; when the study runs online with recruitment-platform integration; when arbitrary video must be played; or when the person building the experiment does not program.

Reach for goxpyriment when the experiment must be distributed as a binary that runs on a machine you do not control and cannot install software on; when timing determinism matters more than stimulus breadth (rapid serial presentation, tight audio-visual synchronisation, hardware-triggered acquisition); when Python environment drift has been a recurring problem; or when you are already comfortable in Go and want the compiler to catch design errors before the participant arrives.

The two are not mutually exclusive. A common pattern is to prototype in PsychoPy and re-implement the timing-critical parts in goxpyriment once the design is settled.

See also