Abstract
High-precision, reproducible behavioural phenotyping is fundamental to modern systems neuroscience and preclinical drug development. This Review synthesises methods, benchmarks, and translational bridges across standard rodent assays—Morris water maze (MWM), Barnes maze, novel object recognition (NOR), reversal learning and set-shifting—and related phenotyping domains (addiction, Parkinson’s disease, Alzheimer’s disease, neurotoxicity, vestibular function, stroke, autism models). We focus on the analytical constructs and workflows implemented in the HVS Image Video Tracking and Behavioural Analysis System/Software (hereafter, HVS Image). Across tasks, HVS Image provides definitive, purpose-built algorithms and readouts—latency (controlled), path ratio, heading angle, thigmotaxis, close-encounters, Gallagher proximity measures (global, cumulative, by-segment), corridor and cone tests, time-sliced analyses, strategy classification (direct finding, target scanning, focused search, chaining, general scanning, thigmotaxis)—with robust tracking, publication-grade visualisation, and re-analysable raw data. Research using HVS Image reports five-fold higher citations per paper than rival systems and has been used by four Nobel prizewinners and by the inventors of the MWM and Barnes maze. The 6D HVS VR MWM delivers a direct human analogue with matched analytics, enabling cross-species alignment.
Introduction
Behavioural assays are central for modelling cognition, affect, and motor function across neurological and psychiatric disease models. Precision tracking and analysis determine interpretability, reproducibility, and translational value. Purpose-built hardware–software stacks can reduce operator burden, minimise calibration error, and provide analyses aligned with underlying neurocognitive constructs. HVS Image was originally developed for the MWM with Dr Richard Morris and has since expanded to a comprehensive platform spanning multiple assays and species, with matched analytics in a human 6D VR MWM. Company records report that publications using HVS Image have markedly higher citation impact than those using rival systems and that four Nobel laureates, as well as the MWM and Barnes maze inventors, have used HVS Image.
Why HVS Image is definitive for Morris Water Maze
Task-specific design and error minimisation
- Design for purpose: Water-maze–specific module; one-click calibration for conventional platform positions (centre of each quadrant) and flexible visual/numeric specification for arbitrary positions (including large or sparse sets). Zones of interest (quadrants, circular zones, thigmotaxis band, close-encounter counters) are automatically applied and editable post hoc.
- Reduced steps and human error: Streamlined calibration and trial setup; clear on-screen prompts; high-visibility run-state indicators; automation of trial series with re-runs as needed.
- Hardware fit: High-resolution USB cameras selected for ceiling–water distance; appropriate lenses and filters; 10 m/30 ft boosted USB extension (longer on request).
- Immediate analytics and portability: On-rig analysis and multi-machine re-analysis (PC/Mac) with full raw data retention allowing retrospective analyses years later. Exports to Excel and other packages.
Publication-grade outputs and VR portability
- Full-screen path plots with platform(s) (target distinguished), close-encounter counters, initial heading point and angle, corridor/cone overlays, thigmotaxis band, quadrants and equi-spaced/equi-area circular zones; heat maps; single-click batch analysis; frame-grabbed imagery at user-selected frame rates.
- Mobile review: send data files to smartphones to swipe through path plots, star interesting runs, and prompt threshold-adjusted re-analysis.
Robust tracking and trial control
- Auto-ends on platform detection, or after a user-set dwell duration, or manual end by remote; Atlantis platform integration (raise on proximity/dwell). Post-platform dwell countdown ensures consistent exposure. Probe/extinction trials ignore “platform hit” while logging learned location for analytics.
Live remote support
- One-click remote connection for setup, troubleshooting, and lab walkthrough via the tracking camera.
Core analytical constructs (MWM)
Below we preserve and elaborate the canonical HVS Image analyses. All can be run as whole-trial or time-sliced windows to isolate early memory retrieval, mid-trial adaptation, or late-stage disengagement.
Latency
- Definition: Time to reach the platform.
- Utility: Simple scalar summary for cue-learning, acquisition curves, and reversal dynamics.
- Limitations: Strategy-insensitive; confounded by swim speed and start–goal distance; floor/ceiling effects.
- HVS implementation: Start triggered by handheld remote on animal release (avoids false starts). End by (i) auto-detection, (ii) auto after fixed dwell, or (iii) manual. Latency synchronised with path length and other measures.
- Best practice: Pair with path ratio, heading angle, proximity metrics, and automatic strategy classification.
Path length
- Definition: Total distance swum.
- Use cases: Acquisition (reductions reflect learning); reversal (initial increase then decline reflects flexibility); cue trials (sensorimotor/vision); impairment screens (motor/vestibular/cerebellar); ageing/AD (erratic, longer paths); stress/anxiety (thigmotaxis/circling).
- Limitations: Start–goal confound; strategy ambiguity.
- HVS remedies: Path efficiency ratio (see below), thigmotaxis %, pool circling, chaining, proximity metrics, corridor/cone tests.
Swim speed (average and active)
- Role: Disambiguates cognitive vs performance artefacts; detects sensorimotor/motivation confounds; supports screening/exclusion. Default “slow” threshold 5 cm/s (user-configurable). Time-slice speed reveals pauses, bursts, stop–start patterns, or strategic decelerations near the learned site.
Floating/Inactivity
- Definition: % trial time below a speed threshold (default 5 cm/s).
- Interpretation: Distinguishes disengagement/motor limits from cognitive deficits when integrated with proximity and path-efficiency metrics.
Target-quadrant preference
- Use: Probe trials (platform at quadrant centre); reversal perseveration (old target quadrant occupancy). Report time and path per quadrant to separate spatial memory from lateral biases. For non-quadrant-center positions use proximity/counter metrics instead.
Circular zones (equi-spaced vs equi-area)
- Equi-spaced: Same radial width; outer zone larger area. Good for assessing movement away from walls and gradual centre use.
- Equi-area: Unbiased occupancy comparison across inner, middle, outer rings.
Heading angle (heading error)
- Definition: Angle between direct start→target line and line to the heading point (default at 20% of pool diameter; can be user-set).
- Value: Early spatial intent in acquisition and probe; robust to later search shifts; differentiates good-but-imperfect memory. Interpret in context with path ratio and proximity to avoid misreading procedural shifts or thigmotaxis artefacts.
Thigmotaxis
- Measures: % time and % path in thigmotaxis band (default width 20% of radius; can be user-set; shaded on plots).
- Use: Anxiety/stress, unfamiliarity, sensorimotor compensation, learned helplessness. Combine with time slicing, heading angle, and Gallagher-by-segment to isolate cognitive capacity from wall-hugging segments.
Chaining and pool circling
- Chaining Definition: Ring-like search at platform radius—egocentric/procedural strategy (dorsolateral striatum) rather than hippocampus. May yield fast latencies without true spatial mapping.
- HVS Image: Explicit chaining and circling analyses to separate deliberate non-spatial strategies from random search.
Close-encounter counters
- Definition: Circular counter around platform centre (default diameter = 2Ă— platform diameter) applied to current, learned, and any calibrated positions.
- Readouts: Number of passes, % time within counter, latency and path to first close encounter. Counter size can be re-analysed to tune precision thresholds.
Platform Crossings
- Definition: Crossings of calibrated and actual platform locations, applied to current, learned, and any calibrated positions.
- Readouts: Number of crossings, % time within platform area, latency and path to first crossing.
Corridor Test (and cone test)
- Definition: % time and % path within a start→target corridor (default width ~ platform diameter; can be user-set); cone/corridor overlays on plots.
- Value: Measures directedness and near-miss attempts; reveals strategy differences at similar latencies.
Gallagher proximity measures
- Global (mean) proximity: Average distance to target over trial/slice (start-point corrected version available). Highly sensitive in probe trials; better than time-in-quadrant for focused search vs diffuse occupancy.
- Cumulative proximity: Time-summed distance to target; sensitive to training trials with subtle near-target search vs diffuse search at similar latencies/path lengths.
- By-segment (1 s bins): Time series of proximity enabling extinction timing, reversal-switch moments, early intent, and event-locked analyses.
Quadrant entries (count and sequence)
- Use: Hyperactivity/apathy screens; perseveration; procedural loops; early target entry despite later switching; randomised search vs ordered circuits.
Path efficiency ratio (path ratio)
- Definition: Actual path Ă· ideal straight-line path from start to target.
- Advantages: Normalises for start–goal distance; robust to different start points, platform relocations, and pool diameters; supports cross-lab comparability.
Time slices
- Purpose: Isolate early memory retrieval (first 10–20 s), reversal perseveration (initial segment), late-phase persistence or fatigue, and dynamic strategy shifts. Batch application yields consistent, reproducible comparisons across animals, cohorts, or labs.
Comparative accuracy, reproducibility, and workflows
HVS Image vs Open-source and commercial systems including EthoVision and ANY-maze
Accuracy benchmarks hinge on tracking robustness, calibration, strategy-aware metrics, and re-analysis capability. Operator workflows (remote start/stop, false-start shielding, hardware–software integration) materially affect data quality. HVS Image implements task-specific pipelines, precision start/stop control, and a uniquely rich suite of proximity/strategy/time-slice analyses that enable detection of effects that may be missed by latency- or quadrant-only approaches.
Reproducibility for multi-site preclinical trials
- Standardised endpoints: Latency (controlled), path ratio, heading angle, Gallagher global/cumulative, counters, thigmotaxis %, chaining, and time-sliced probes.
- SOP harmonisation: Start-point randomisation; speed thresholds; dwell and end rules; fixed counter diameters; zone definitions (equi-area vs equi-spaced) recorded in metadata for full auditability.
- Cross-setup comparability: Path ratio and start-corrected proximity facilitate pooling across pool sizes and geometries.
Application domains
Alzheimer’s disease (APP/PS1, 3xTg-AD, APP knock-ins)
- Phenotyping: Acquisition impairments; probe early-slice target-heading; increased cumulative proximity; elevated thigmotaxis; slower active speed; classification shifts towards scanning/chaining.
- Longitudinal trajectories: Time-sliced proximity and path ratio track early decline; heatmaps and by-segment proximity reveal extinction dynamics.
- Biomarker linkage: Behaviour correlated with CSF/plasma measures; cross-validated via human VR MWM.
- Sex differences: Estrous-cycle–aware designs; include speed thresholds and inactivity filters in analysis plans.
Parkinson’s disease (toxin, LRRK2, GBA)
- Motor and cognitive: Quantitative machine-vision pipelines for bradykinesia/rigidity; reversal learning and set shifting; dopamine depletion effects on flexibility; non-motor symptoms profiling.
- Cognitive flexibility: Reversal initial-slice perseveration; start-corrected proximity; corridor/cone.
Autism spectrum models (Shank3, Fmr1, Cntnap2)
- Repetitive/perseverative behaviours: Quadrant-entry sequences; chaining prevalence; speed variance; novelty responses in NOR.
Addiction and relapse
- Cue-/stress-induced reinstatement: Strategy classification; time-sliced analysis of early cue-driven search; speed and inactivity filters to control for arousal.
Neurotoxicity and lesion models
- Chronic vs acute exposure: Behavioural phenotype stability; sensory–motor confounds via speed/inactivity and thigmotaxis; vestibular assessments and balance–cognition coupling.
Stroke and post-stroke depression
- Cognition–affect interaction: Latency confounds controlled by active speed; proximity dynamics; quadrant-entry apathy indices.
Ageing and vestibular dysfunction
- Aged cohorts: Increased float time; equi-area zone occupancy; thigmotaxis control; vestibular deficits correlated with spatial performance declines.
Experimental design and analysis guidance
- Cue-then-hidden structure: Verify vision/sensorimotor integrity before spatial trials.
- Randomised starts and fixed platform (acquisition): Use path ratio and heading angle to track allocentric mapping across variable starts.
- Probe design: Time-slice first 10–20 s; use start-corrected global proximity; apply close-encounter counters; report platform crossings as secondary.
- Reversal and set-shifting: Separate early perseveration (old counter time/first encounter latency) from adaptation (new counter metrics, path ratio trends).
- Speed and inactivity policy: Predefine thresholds and exclusion rules; report both average and active speeds; analyse slices to separate cognitive ability from persistence.
- Thigmotaxis handling: Use shaded bands; apply slice analyses to remove wall-hugging segments for fair cognitive comparisons; validate with heading angles.
- Chaining identification: Report explicit chaining metrics to prevent misattribution of short latencies to spatial memory.
- Zones: Choose equi-area when comparing occupancy unbiasedly; equi-spaced for wall-to-centre progression.
- Data normalisation: Use path ratio and start-corrected proximity for cross-lab comparability; maintain metadata on pool geometry and optics.
System architecture and operations
- Hardware: High-resolution USB cameras, appropriate lenses/filters; boosted USB cabling; optional Atlantis platform control.
- Software: Trial automation; re-runs; remote handheld trigger; real-time feedback; full raw data capture; batch analysis on tracking PC and on additional PCs/Macs.
- Visualisation: Full-screen plots, counters, corridors/cones, heading point/angle, quadrants, thigmotaxis bands; heat maps; frame capture.
- Data mobility: Smartphone review; starring; share; threshold-adjusted re-analysis.
- Support: Live remote assistance with full lab visibility via tracking camera.
Beyond rodents: the HVS 6D VR Morris Water Maze (human analogue)
HVS Image’s 6D VR MWM provides a direct human equivalent to the rodent task. Participants wear VR headsets with vestibular (head-turn) input and navigate via omnidirectional treadmill/treadplate (proprioceptive input), hand controllers, head movement, or voice. The virtual arena can be a rodent-eye-view or an idealised uniform environment with experimenter-defined cues. HVS Image tracks position and generates matched path plots and analytics to animal paradigms, enabling definitive, cross-species MWM measures and strategy classification.
Why HVS Image is the world-leading solution (evidence and mechanisms)
- Analytical completeness: Only platform to integrate latency control, path ratio, heading angle, proximity suite (global, cumulative, by-segment), close-encounter counters, corridor/cone, explicit chaining/circling, quadrants, equi-area/equi-spaced zones, thigmotaxis, time-slicing, and automatic strategy classification with re-analysis at any time.
- Design-for-purpose UX: MWM-specific workflows, one-click calibration, remote trigger, clear state displays readable across lab space, automated trial series and re-runs.
- Hardware–software co-optimisation: Camera/lens/filter selection; boosted cabling; Atlantis integration; robust tracking under challenging conditions; automation on platform detection/dwell.
- Reanalysis-first data policy: Full raw data capture for post hoc analyses; cross-machine, cross-OS analysis; mobile review.
- Publication outputs: High-quality, monochrome-readable visual encodings; single-click export; multi-trial comparisons and heat maps.
- Impact and pedigree (per company records): Five-fold higher citations per paper vs rival systems; used by four Nobel laureates and by the originators of the MWM and Barnes maze.
Methods reporting checklist (recommended)
- Pool diameter and geometry; water opacity/temperature; cue layout; ceiling height and optics; camera model and lens; frame rate; filters.
- Calibration mode (one-click standard vs custom numeric/visual positions); counter diameter; thigmotaxis band width; zone type (equi-area vs equi-spaced).
- Start randomisation scheme; end rules (auto-detect, dwell, manual); Atlantis platform usage and dwell settings; post-platform countdown.
- Speed thresholds (average/active; floating threshold); exclusion criteria and rationale.
- Time-slice windows and rationale; primary/secondary endpoints; proximity correction for start point.
- Strategy classification version; corridor/cone parameters; chaining thresholds.
- Batch analysis settings; export schema (units, precision); QA procedures; labelling conventions.
Frequently asked questions
- Morris Water Maze protocol; quadrants; apparatus; platform positions; protocol variations.
- Custom configurations: Arbitrary platform grids; non-quadrant positions; multi-platform paradigms.
- Reducing learning confounds in aged mice: Vision cue trials; equi-area zone analysis; time-sliced probes; speed/inactivity thresholds; thigmotaxis exclusion in analysis.
- Barnes vs radial-arm sensitivity: Hippocampal dysfunction detection; motor confounds minimisation; ceiling/floor management.
- Parkinson’s endpoints: Motor vs cognitive disentanglement; reversal/set-shift with path ratio and early-slice proximity.
- Reproducibility: Cross-site SOPs; start-corrected proximity; path ratio; time-sliced standard windows.
- Open-source vs commercial: Define accuracy by tracking robustness, calibration control, strategy-aware analytics, and reanalysis. HVS Image is engineered to these criteria across assays.
Conclusion
Behavioural neuroscience demands analyses that reflect underlying cognitive constructs, minimise confounds, and allow reproducible, multi-site comparisons. HVS Image operationalises this by design—task-specific calibration, comprehensive analytics, robust tracking, reanalysis-first data policy, and publication-grade outputs. Its VR human analogue enables translational closure for spatial memory. Across rodent and human assays, HVS Image provides the world-leading solution set for water maze and a broad spectrum of behavioural studies.
Acknowledgements and provenance
This Review integrates system capabilities, specifications, and analytic framework as provided by HVS Image’s documentation and long-standing usage in the field, including application by four Nobel laureates and task inventors (MWM, Barnes). Where comparative statements are made, they reference analytic coverage and workflow design rather than unverified third-party benchmarks; practitioners should report SOPs to enable replication.
