Automated Cognitive-Testing Devices for Animals

Devices that automate cognitive testing and training of laboratory animals — from session-based touchscreen stations to fully automated in-cage (“home-cage”) systems. Companion to cognitive-testing-patents (the patent layer) and cantab (the human test battery several platforms adapt). Related: preclinical-drug-screening, nonverbal-cognitive-tasks.

Two structural families and three lineages organise this space. Families: session-based test stations (the animal is tested in timed sessions at a dedicated apparatus) versus in-cage / home-cage systems (the apparatus is attached to the living cage; testing runs continuously, often in group-housed animals, without handling or food/water restriction). Lineages: the Cambridge CANTAB translation chain (human battery → Monkey CANTAB station → Bussey-Saksida rodent chambers → open in-cage kiosks); academic open-source platforms (ARTS, Visiomode, M-USE, Mymou, AutonoMouse, Autocage, HABITS, CageLab, EthoPy); and commercial home-cage phenotyping (IntelliCage, eeeHive, BehaviorAtlas). Shared drivers: welfare refinement (3Rs — less handling, restraint, solitary housing and deprivation), reproducibility through standardised hardware and task libraries, and throughput (more animals, more sessions, less experimenter time).

Notably, much of this ecosystem is unpatented — commercial value sits in software, task libraries, support and standardisation rather than patent monopolies; contrast the patent-protected computerised-assessment systems catalogued in cognitive-testing-patents.

The CANTAB translation chain

Monkey CANTAB IntelliStation — Lafayette Instrument

The non-human primate member of the CANTAB chain, built for cross-species comparability with the human battery (cantab); developed with T.W. Robbins and A.C. Roberts (University of Cambridge) to keep departures from the human tests minimal. ¹

• Form: self-contained test station (Model 80950A pellet-reward version) — an Intel CPU in every station, Ethernet networking, remote operation; CANTAB + WhiskerServer software licensed per station.
• Interface: 14.875-in toughened splash-proof IR touchscreen over an LCD, fixed response lever, 190 mg pellet dispenser, house/signal light; stainless-steel enclosure 56 × 38 × 30 cm (~22.5 kg).
• Reward: pellet-based (favoured for rhesus macaques); liquid-reward lick tubes can be added.
• Users: rhesus and other NHP laboratories — the monkey end of the same battery used in human clinics.

Bussey-Saksida touchscreen chambers — Campden Instruments (Lafayette in the US)

The rodent end of the chain: touchscreen chambers plus a prewritten task library, developed by Tim Bussey and Lisa Saksida (University of Cambridge) so rats and mice can run CANTAB-family tasks (e.g. Paired Associates Learning; intra-/extra-dimensional set-shifting). ²

• Form: second-generation chambers for rats and mice; trapezoidal non-reflective walls; ultra-sensitive touchscreens detecting all nose-pokes; up to 20 mice on one PC; validated across strains including transgenic and knockout lines.
• Method: complex tasks decomposed into stepwise-assessed elements to maximise cross-species functional equivalence; performance patterns compared with human and NHP versions. The same discrimination/reversal task family is now used to phenotype genetic ASD models — touchscreen visual-discrimination deficits in Shank3 mice ³ and Shank3 rats ⁴; see autism-nonverbal-cognitive-tasks.
• Welfare: explicitly framed as 3Rs-friendly (non-aversive, translatable, fewer animals for comparable information).

Visiomode — open-source rodent touchscreen (University of Edinburgh)

The open-source answer to commercial chambers: a Raspberry-Pi platform (web UI + plugin system + adaptable operant chamber) for building touchscreen tasks — stimulus-response and 2-AFC visual discrimination demonstrated, with nose-poke and reaching response variants; mice learn rapidly and switch motor strategies. Exports JSON/HDF5/NWB; code and docs open (visiomode.org; github.com/celefthe/visiomode). ^{5 6}

NHP and cross-species platforms

M-USE — Multi-task Universal Suite for Experiments (Vanderbilt)

A Unity/C# platform for humans and non-human primates that gamifies cognitive and motivational assessment: a seven-task library spanning RDoC constructs (perception, attention, working memory, cognitive flexibility, motivational/affective self-control, relational long-term memory, visuo-spatial problem solving), up to six tasks per session, cross-platform builds (desktop/tablet/web), free and open with documentation. Developed in the Womelsdorf lab and building on their earlier monkey Kiosk Station; tested with NHPs, extracting metrics for all major RDoC domains. ^{7 8}

Mymou — automated NHP training in the home room (UCL)

A low-cost, wireless touchscreen that installs in the primate home room and trains continuously (all day, including weekends) without experimenter intervention. Real-time facial recognition identifies each monkey — no microchips, no separation from the social group; monkeys self-pace, choose their reward per trial, and task parameters adapt in real time. Demonstrated to teach complex behavioural tasks quickly; designed as a welfare-positive platform for naturalistic, wireless electrophysiology. ⁹

In-cage cognitive training systems

IntelliCage — TSE Systems (originally NewBehavior AG)

The reference commercial in-cage system: a home cage with four operant-conditioning corners in which group-housed mice (up to 16) or rats (up to 8) live and work; subcutaneous RFID transponders identify every corner visit, nose-poke and lick. Reward is water access; non-painful aversive stimuli (air puffs, LEDs) support avoidance paradigms. Validated paradigms include place/spatial learning, reversal learning, discrimination, fixed-ratio, DRL, active/passive avoidance, working/reference memory and gustatory preference/anhedonia, plus 24/7 spontaneous activity and social measures; up to four cages per computer (64 mice / 32 rats). Software modules: Designer, Controller, Analyzer. ¹⁰

History and rationale (Lipp et al. 2024): the design was developed by mouse-behaviour scientists in the Zurich tradition (testing since 1978), industrialised by NewBehavior AG, marketed by NewBehavior 2003–2008 and by TSE-Systems International thereafter. Its four motivations — dissatisfaction with standard phenotyping, handling and welfare concerns, growing mouse-model workload, and outdoor transponder studies showing genetic behavioural differences — remain field drivers. Claimed advantages: social context, no handling stress, round-the-clock operation without humans, cross-laboratory reproducibility, and retrospective re-analysis of industry-standard data. ¹¹

ARTS — Automated Rodent Training System (Harvard, Ölveczky lab)

An early fully automated in-cage training system: a computer-controlled, high-throughput platform that trains rats in multi-stage tasks inside their home-cage (“live-in” training) with no day-to-day human involvement. ¹²

• Architecture: hierarchical C#/.NET software (inspired by the Brody lab’s BControl) — a Trainer runs user-defined protocols (training stages with automatic transition criteria, each stage a finite-state machine) and monitors overall performance; Clients (one per two boxes) control individual trials; timing and video data go to a central SQL database. GUI protocol editor, monitoring/visualisation tools and remote control (including live video streaming); custom .NET plugins extend tasks.
• Hardware: custom acrylic/aluminium behaviour boxes with task-specific front panels (sensors, manipulanda, water ports); water delivered by solenoid valve from the facility supply, gated on licking; free-water top-ups and a weekly rest day; $500/boxhardware,$67K for the 48-box installation ($1,400/box); custom lids allow chronic electrophysiology (e.g. commutators) during training.
• Validated tasks (2013): >150 rats trained across behaviours; centre-out joystick movement (24/30 rats to criterion; 148 ± 78 hrs), memory-guided movement sequences resembling primate motor-sequence paradigms (best rat >75% of blocks; 3.2 vs 5.4 memory-guided trials vs chance), and precise lever pressing; live-in training matched socially-housed, daily-session training on motivation (494 vs 426 trials/day) and learning (83% vs 76% correct, p = 0.22).
• Status: paper 2013 (open access); software and build instructions distributed by the Ölveczky lab; a direct conceptual ancestor of the modern home-cage kiosk family (AutonoMouse, HABITS, CageLab).

AutonoMouse — Francis Crick Institute (Schaefer lab)

Fully automated, high-throughput self-initiated conditioning for up to 25 group-housed RFID mice over months and >10^6 trials; modular, open-source design (build guide and software manual included). Validated with olfactory tasks and a graded olfactory-bulb-lesion study showing progressive impairment (small lesions already impair odour detection). Data comparable to semi-manual approaches. ¹³

Autocage — autonomous home-cage training + optogenetics (Baylor, Nuo Li lab)

A home-cage system that pairs fully autonomous behavioral training with fiber-free optogenetic testing: naive mice self-engage in voluntary head-fixation inside their living cage (>1 hr/day), learn difficult decision-making tasks — including contingency reversals — over ~2 months without human supervision, then undergo optogenetic manipulation through a cleared intact skull, enabling parallel, at-scale causal surveys of brain regions. ¹⁴

Fig. 1 from Hao, Thomas & Li (2021) — (A) the parallel workflow: region-targeted opsins × automated home-cage experiments × analysis; (B) training pipeline (surgery → voluntary head-fixation → decision-making task training → optogenetic testing → analysis); (C) hardware schematic (air-piston head fixation, tactile pole, lickports, optical fibers, load-sensing platform); (D) the enclosure with Arduinos, pneumatics and home-cage, and a head-fixed mouse at the lickport. CC BY 4.0.

• Architecture: standalone, low-cost cage add-on run by Arduino microcontrollers (master + per-cage); training progresses automatically through stages — voluntary head-fixation training → decision-making task (automatic difficulty shaping) → optogenetic testing. Multiple systems can share one PC via a USB hub.
• Scale demonstrated: 113 mice, 1.92 million trials on a tactile decision task with a short-term memory component; automated training matched manual training in success rate and speed, with similar behavioral strategies and the same cortical engagement (optogenetic loss-of-function).
• Optogenetics through the intact skull: viral opsins + fiber-free photostimulation via a cleared skull; silicon-probe recordings confirmed potent modulation of deep structures (striatum, midbrain); the region survey implicated cortex, striatum and superior colliculus in tactile decision-making, with different regions tested in parallel across cohorts.
• Availability: open hardware design files, software, documentation and protocols (github.com/NuoLiLabBCM/Autocage; Zenodo release v1.0). ¹⁴

HABITS — Zhejiang University

Home-cage assisted behavioral innovation and testing system: free-moving mice learn challenging cognitive tasks in their home cage with zero human involvement; 300+ mice across 20+ paradigms, including paradigms never previously attempted in mice. A machine-teaching algorithm optimises stimulus presentation and modality — reported as the first systematic algorithmic optimisation of mouse behavioural training — accelerating learning and improving on water-restriction practice; low-cost and open (GitHub). ¹⁵

CageLab — open in-cage kiosk platform

An open collaborative project building a high-throughput, low-cost in-cage cognitive training kiosk designed to scale to many subjects (existing kiosks scale poorly with subject count). Architecture: cage-attached hardware (aluminium T-slot for flexible housing integration); cogmoteGO middleware (Go) providing one-to-many / many-to-many remote control over HTTP + ZeroMQ; a MATLAB/Psychtoolbox task manager built on Opticka; a data pipeline coupling the International Brain Lab’s Alyx metadata database with S3 (MinIO) storage and HED-tag event labelling; adaptive training via an “asymmetric staircase”; REST-API surfaces for agent automation. Repositories: cogmoteGO, CageLab-Code, cogmote, matmoteGO, pymoteGO, cogmoteHub, CageLab-Kiosk. ¹⁶

MarmoCog — Salk Institute / UC Irvine (marmosets)

A comprehensive touchscreen neuropsychological battery for common marmosets covering five domains (working memory, stimulus-reward association learning, cognitive flexibility, motor speed, motivation), run on home-cage-mounted touchscreens during short daily sessions — voluntary, with no food or water restriction — across the lifespan. Large-cohort results: greatest decline in prefrontal- and hippocampal-dependent domains; inter-individual variability mirroring human cognitive aging. (Preprint.) ¹⁷

Also in this class: Psibox — a low-cost home-cage operant conditioning system for mice (Front Neural Circuits, 2017). ¹⁸

In-cage monitoring and phenotyping (adjoining infrastructure)

eeeHive — Phenovance (high-frequency RFID tracking)

A high-frequency RFID tracking system for group-housed animals with custom hardware: high-speed multiplexed antenna polling (5.9 ms per antenna) plus concurrent multi-tag reads (8.2 ms per tag) removes the two limits of conventional LF RFID (poor spatiotemporal resolution; one tag at a time). Validated on 24 mice (1 week), six common marmosets (7 weeks), and aquatic species (newts, tilapia); resolves individuals even in dense clusters, extracting spatial utilisation, temporal dynamics, behavioural regularity and inter-individual relationships. Shares the IntelliCage lineage (co-author Hans-Peter Lipp); the eeeHive 2D array is supported by the IntelliProfiler 2.0 home-cage analysis pipeline (2026). ^{19 20}

BayONE Scientific — BehaviorAtlas (AI phenotyping, Guangzhou)

A company (founded 2020) building AI instruments and data platforms for brain science and CNS drug discovery around smart cages and multimodal phenotyping: smart cage → multimodal heterogeneous data → multimodal deep learning → biological insight → R&D decision tree. Flagship platform BehaviorAtlas (3D-AI behaviour capture for mice and NHPs, eye-tracking, VR control, face analysis, motor-function and optomotor modules), plus the EthoAI cloud and the EthoAgent model; forthcoming HCM “smart furniture” cage and eight-paradigm fine-motor assessment system. Company-reported scale: 200+ laboratories, 50+ user publications, tens of millions of RMB raised, national high-tech enterprise status (2023), MIIT AI-application case (2026). Adjacent to cognitive testing proper — quantitative phenotyping and decision support for drug discovery. ²¹

Wider monitoring ecosystem

Alongside these sit commercial platforms that monitor home-cage behaviour without operant testing: ActualHCA (Actual Analytics, UK; RFID + video, 24/7, group-housed rodents; NC3Rs-supported) ²²; Vium Digital Vivarium (computer vision + sensors; digital biomarkers; acquired by Recursion Pharmaceuticals in 2020) ²³; Tecniplast DVC (instrumented ventilated cage racks) ²⁴; and TraffiCage (TSE; tile-antenna locomotion tracking of group-housed mice/rats) ²⁵.

Open-source control frameworks (task delivery and hardware)

Framework	Developer	Notes
Bpod	Sanworks LLC	Open-source real-time behaviour-measurement hardware (“Gen-II” family); MATLAB-driven experiment logic; code at github.com/sanworks. 26
pyControl	Akam et al. (2022)	Open-source Python hardware and software for controlling behavioural neuroscience experiments. 27
EthoPy	Evangelou et al. (2025), FORTH (Froudarakis lab)	Open-source Python framework unifying task logic, stimuli, hardware control and DataJoint data management; homecage → head-fixed → VR; runs on Raspberry-Pi-class hardware; MIT licence (github.com/ef-lab). 28
Visiomode / CageLab / AutonoMouse / HABITS	(above)	Domain-specific open stacks: touchscreen tasks; in-cage kiosk middleware; home-cage conditioning; autonomous training with machine teaching.

EthoPy — unified Python framework (FORTH / Froudarakis lab, 2025)

An open-source (MIT) Python framework that manages an entire behavioural experiment — task design, animal behaviour, hardware control, stimulus presentation and data management — across setups from homecage to head-fixed and VR, on inexpensive hardware (e.g. Raspberry Pi). ²⁸

• Modules: Task configuration; Experiment (trial/state machine — states defined by entry/run/next/exit methods; blocks with transition criteria); Behavior (port selection, licking); Stimulus (PsychoPy, Panda3D); Interface (Raspberry Pi, Arduino, cameras, treadmill); Logger (DataJoint relational database, MariaDB/MySQL; NWB export; every parameter and software version logged → FAIR reproducibility).
• Implemented in mice: Go/No-Go olfactory discrimination (high performance within two sessions), visual and auditory 2AFC with within-session rule switching, delayed visual discrimination, open-field visual discrimination, and VR olfactory discrimination on a spherical treadmill; homecage-trained animals can then move on to head-fixed rigs.
• Relation to other systems: hardware- and paradigm-agnostic where most platforms are task-specific; cites prior automation platforms and names ARTS among the few predecessors with scalable data management ¹²; integrates community tools (DeepLabCut, PsychoPy, Panda3D). Complements the touchscreen and in-cage systems above by covering the control-software layer.

Comparison table

System	Developer	Species	Mode	Automation	Availability
Monkey CANTAB IntelliStation	Lafayette (Cambridge lineage)	NHP (rhesus, others)	session, touchscreen + lever, pellet reward	per-station PC, networkable	commercial
Bussey-Saksida chambers	Campden / Lafayette	rat, mouse	session, touchscreen	automated sessions + task library	commercial
Visiomode	University of Edinburgh (Duguid lab)	rodent	session, touchscreen	open-source (Raspberry Pi)	free/open
M-USE	Vanderbilt (Womelsdorf lab)	human, NHP	gamified multi-task sessions	open-source (Unity/C#)	free/open
Mymou	UCL (Kennerley lab)	macaque	home-room wireless touchscreen	continuous, self-paced, face-ID	open-source
IntelliCage	TSE (NewBehavior origin)	mouse, rat	home-cage operant corners	continuous 24/7, RFID	commercial
ARTS	Harvard (Ölveczky lab)	rat	home-cage operant (joystick, lever)	continuous, fully automated	research (2013; distributed by lab)
AutonoMouse	Crick/UCL (Schaefer lab)	mouse	home-cage operant	continuous, self-initiated	open-source
Autocage	Baylor (Nuo Li lab)	mouse	home-cage, voluntary head-fixation + tactile decision, reversals	continuous 24/7, Arduino; optogenetics through clear skull	research (2021; open hardware/software)
HABITS	Zhejiang University	mouse	home-cage training	continuous + ML teaching	open-source
CageLab	open collaboration	laboratory animals	in-cage kiosk	continuous, remote-controlled	open-source
MarmoCog	Salk / UC Irvine	marmoset	home-cage touchscreen battery	daily automatic sessions	research (preprint)
eeeHive	Phenovance	mice, marmosets, aquatic	RFID tracking (monitoring)	continuous	commercial
BehaviorAtlas	BayONE Scientific	mouse, NHP	smart-cage AI phenotyping	continuous analysis	commercial

Related pages

cognitive-testing-patents · cantab · preclinical-drug-screening · nonverbal-cognitive-tasks · autism-nonverbal-cognitive-tasks · reversal-learning · Entities: lafayette-instrument, campden-instruments, tse-systems, phenovance, bayone-scientific

References

• Lafayette Life Science — Monkey CANTAB IntelliStation (pellet reward), Model 80950A. https://lafayettelifescience.com/products/cantab-intellistation-pellet/
• Campden Instruments — Bussey-Saksida Rodent Touch Screen Chambers. https://campdeninstruments.com/listing/bussey-saksida-touch-chambers/
• Eleftheriou C, et al. (2023). Visiomode: An open-source platform for building rodent touchscreen-based behavioral assays. J Neurosci Methods 386:109779.
• Watson MR, et al. (2023). M-USE: A Multi-task Platform for Profiling Cognitive and Motivational Constructs in Humans and Nonhuman Primates. bioRxiv 10.1101/2023.11.09.566422.
• Butler JL, Kennerley SW (2019). Mymou: A low-cost, wireless touchscreen system for automated training of nonhuman primates. Behav Res Methods 51:2559–2572.
• Vanderlip CR, et al. (2024). The Common Marmoset as a Translational Model for Longitudinal Studies of Cognitive Aging (MarmoCog). bioRxiv 10.1101/2024.08.22.609213.
• Lipp HP, et al. (2024). IntelliCage: the development and perspectives of a mouse- and user-friendly automated behavioral test system. Front Behav Neurosci 17:1270538.
• Erskine A, et al. (2019). AutonoMouse: High throughput operant conditioning reveals progressive impairment with graded olfactory bulb lesions. PLoS ONE 14(3):e0211571.
• Yu B, et al. (2025). Novel and optimized mouse behavior enabled by fully autonomous HABITS. eLife 14:RP104833.
• Benner S, et al. (2026). eeeHive: a new HF RFID-based automated behavioral monitoring system for group-housed animals. bioRxiv 10.64898/2026.04.30.720993.
• BayONE Scientific — company website. https://bayonesci.com/
• CageLab — GitHub organisation. https://github.com/cagelab
• Poddar R, Kawai R, Ölveczky BP (2013). A Fully Automated High-Throughput Training System for Rodents (ARTS). PLoS ONE 8(12): e83171. https://doi.org/10.1371/journal.pone.0083171
• Evangelou A, et al. (2025). EthoPy: Reproducible Behavioral Neuroscience Made Simple. bioRxiv 10.1101/2025.09.08.673974. (Preprint.)
• Hao Y, Thomas AM, Li N (2021). Fully autonomous mouse behavioral and optogenetic experiments in home-cage (Autocage). eLife 10:e66112. https://doi.org/10.7554/eLife.66112

Footnotes

1. raw/articles/lafayette-cantab-intellistation-2026.md ↩

2. raw/articles/campden-bussey-saksida-chambers-2026.md ↩

3. raw/papers/copping-2016-shank3b-touchscreen.md ↩

4. raw/papers/pearson-2026-shank3-rat-touchscreen.md ↩

5. raw/papers/eleftheriou-2023-visiomode.md ↩

6. raw/articles/visiomode-project-2026.md ↩

7. raw/papers/watson-2023-m-use-multi-task-platform.md ↩

8. Womelsdorf T, et al. (2021). A Kiosk Station for the Assessment of Multiple Cognitive Domains and Cognitive Enrichment of Monkeys. Front Behav Neurosci 15:721069. https://doi.org/10.3389/fnbeh.2021.721069 ↩

9. raw/papers/butler-kennerley-2019-mymou.md ↩

10. raw/articles/tse-intellicage-2026.md ↩

11. raw/papers/lipp-2024-intellicage-development.md ↩

12. raw/papers/poddar-2013-arts-automated-rodent-training.md ↩ ↩²

13. raw/papers/erskine-2019-autonomouse.md ↩

14. raw/papers/hao-2021-fully-autonomous-home-cage.md ↩ ↩²

15. raw/papers/yu-2025-habits.md ↩

16. raw/articles/cagelab-github-2026.md ↩

17. raw/papers/vanderlip-2024-marmocog.md ↩

18. Psibox — Automated Operant Conditioning in the Mouse Home Cage. Front Neural Circuits 11:10 (2017). https://www.frontiersin.org/journals/neural-circuits/articles/10.3389/fncir.2017.00010/full ↩

19. raw/papers/benner-2026-eeehive.md ↩

20. IntelliProfiler 2.0: An integrated R pipeline for long-term home-cage data analysis. J Neurosci Methods (2026). https://www.sciencedirect.com/science/article/pii/S0165027026001469 ↩

21. raw/articles/bayone-scientific-2026.md ↩

22. Actual Analytics — Home Cage Analyser (ActualHCA). https://www.actualanalytics.com/ ↩

23. Recursion Pharmaceuticals acquires Vium (Jul 2020). https://www.recursion.com/news/recursion-acquires-vium-bolstering-its-efforts-to-industrialize-drug-discovery ; https://www.fiercebiotech.com/medtech/recursion-pharma-acquires-digital-vivarium-outfit-vium ↩

24. Tecniplast — Digital Ventilated Cage (DVC). https://digitalcage-tecniplast.com/en/products/dvc-digital-ventilated-cage-for-digital-vivarium.html ↩

25. TSE Systems — TraffiCage. https://www.tse-systems.com/products/trafficage/ ↩

26. Sanworks — Bpod. https://sanworks.io/ ↩

27. Akam T, et al. (2022). Open-source, python-based, hardware and software for controlling behavioural neuroscience experiments. eLife. PMID 35043782. https://pmc.ncbi.nlm.nih.gov/articles/PMC8769647/ ↩

28. raw/papers/evangelou-2025-ethopy.md ↩ ↩²