How MWM Escape Latency Is Calculated

1. Supported workflow with validation-only single-trial measures

This page documents how RodentMetrics currently computes and reports these measures for one trial -- it does not imply that the metrics, or RodentMetrics' Morris Water Maze analysis generally, have been independently validated.

2. What escape latency represents

Escape latency is the elapsed time from the start of the analyzed interval to the first confirmed moment the animal reached the platform. It is commonly used in the literature as an indicator of how quickly an animal locates an escape platform in a given trial. RodentMetrics reports it in that same descriptive sense -- a measured duration for this one trial, not a judgment about learning or memory.

3. How RodentMetrics defines the pool and platform

The pool is calibrated as an ellipse in the source video (to account for camera angle), which RodentMetrics maps into a true circular pool in physical centimeters using your configured physical diameter. All positions, distances and quadrants used elsewhere on this page are computed in that calibrated, circular coordinate space, not in raw pixel coordinates.

The platform is defined as a circle: a center point plus a diameter in centimeters. For a Probe Trial, that same circle marks where the platform used to be (the platform is physically removed during a probe trial). The animal's position is read from the tracker's own tracked point for each frame, transformed into the calibrated pool coordinates -- not a separate pose-based keypoint.

4. What counts as reaching the platform

A frame counts toward reaching the platform when the tracked point's distance to the platform center is within the platform radius. That alone only starts a candidate reach -- it becomes a confirmed reach only once the animal has remained continuously inside the platform circle for at least 0.20 seconds. A candidate is discarded (and must restart from scratch) if tracking goes missing, the tracked point becomes invalid or leaves the pool, an implausible position jump breaks the trajectory, or the animal exits the platform circle before the 0.20-second window completes -- a brief pass through the platform does not count.

RodentMetrics uses only the first confirmed reach in a trial. Once a reach is confirmed, no later platform visits are evaluated for this metric. The reach's reported time is the moment the animal actually entered the platform circle, not the later moment the 0.20-second confirmation completed -- the confirmation window exists to filter out noise, not to delay the reported time.

5. The exact escape-latency calculation

Escape latency is: the confirmed reach's entry timestamp, minus the analysis start time (the configured start point of the analyzed interval, in seconds -- not necessarily the first frame of the video). The result is reported in seconds and is not rounded at this stage.

6. Worked example

Take the following illustrative, synthetic Hidden Platform trial (not data from a real recording), with continuous tracking assumed throughout the swim to the platform:

RodentMetrics would report an escape latency of 23.4 s, a path length to the platform of 421.2 cm, and a mean swim speed to the platform of 18.0 cm/s for this trial.

7. Path length and swim speed

RodentMetrics reports two distinct pairs of distance/speed measures, not one:

• Path length to platform and mean swim speed to platform -- accumulated only from the analysis start up through the confirmed platform reach. These are unavailable when the platform is never reached, matching escape latency's own availability.
• Whole-trial path length and whole-trial mean swim speed -- accumulated across the entire analyzed interval, independent of whether or when the platform was reached.

Both are calibrated in physical centimeters, not pixels. Distance only accumulates between consecutive, validly tracked frames whose displacement is plausible; a missing frame, a detection outside the pool, or an implausible position jump breaks the path at that point rather than being bridged or interpolated across.

8. Quadrant occupancy and former-platform crossings (Probe Trials)

Quadrant occupancy and former-platform crossing counts are computed only for Probe Trials -- they are not part of a Hidden Platform or Visible Platform result. The pool is divided into four quadrants around its calibrated center, using the same rotatable reference axes you set when defining the pool; the target quadrant is derived automatically from the platform's configured location, not chosen separately. Quadrant time is the count of valid tracked frames in that quadrant, divided by the total valid tracked frames in the trial -- for example, 21.6 s of target-quadrant time in a 60.0 s fully-tracked probe trial would be reported as 36.0%. Frames with missing or invalid tracking are excluded from every quadrant's count, never guessed at.

Separately, RodentMetrics counts confirmed, discrete visits back to where the platform used to be ("former-platform crossings"), each requiring the same style of entry confirmation used for platform reach, plus a matching period spent away from that location before a new visit can be counted -- so one continuous visit is never counted as several.

9. Boundary cases

A few situations are worth understanding before interpreting a result:

• If the animal is already on the platform at the first valid tracking frame, escape latency is unavailable rather than 0 (Section 5).
• If the platform is never reached, escape latency, path length to platform and mean swim speed to platform are all unavailable, not 0 or the trial duration.
• A brief pass through the platform that does not sustain the 0.20-second confirmation window does not register as a reach.
• A candidate reach interrupted by missing tracking, an implausible position jump, or the animal leaving the pool boundary is discarded and does not carry over -- a fresh candidate is required afterward.
• With zero valid tracked frames anywhere in the trial, every position-derived measure on this page is reported as unavailable due to insufficient tracking evidence, not estimated.

10. What these metrics do not establish

Escape latency, path length and quadrant/crossing measures describe this one trial's tracked swim path. On their own, they do not:

• independently prove spatial learning took place;
• diagnose or measure a cognitive impairment;
• substitute for a repeated-trial experimental design -- a single trial is one observation, not an acquisition curve;
• demonstrate that RodentMetrics' own Morris Water Maze implementation has been independently validated.

They should be interpreted alongside tracking/quality information for the trial and the researcher's experimental protocol -- not read in isolation as a conclusion about the animal.

11. What to report

Because the current Morris Water Maze measures are validation-only and single-trial, RodentMetrics recommends reporting the underlying parameters alongside any single metric:

• Escape latency, or that the platform was never reached
• Whether the platform was reached, and the platform geometry used
• Path length and mean swim speed, both to-platform and whole-trial
• Quadrant occupancy and target quadrant, for Probe Trials
• Former-platform crossing count, for Probe Trials
• Tracking/quality information for the trial
• That results come from a single trial, not a multi-trial acquisition curve
• The RodentMetrics software version the analysis was run with

Further reading

Vorhees CV, Williams MT. Morris water maze: procedures for assessing spatial and related forms of learning and memory. Nat Protoc. 2006;1(2):848-858. doi:10.1038/nprot.2006.116