Morris Water Maze Test

Abstract

The Morris water maze test, which MeSH classifies under maze learning, is a behavioral task in which a rodent swims in opaque water to locate a platform hidden just beneath the surface, using only the distal cues around the room. Because the platform cannot be seen, smelled, or felt until it is reached, efficient escape requires the animal to learn its location relative to the environment — a form of place learning dependent on the hippocampus. Devised by Richard Morris in the early 1980s, the task became the standard assay of spatial memory after hippocampal lesions were shown to abolish place navigation while sparing the ability to swim to a visible platform. This article describes the apparatus and protocol, the probe trial that indexes spatial memory, the measures used to score it, and the task's modern computational refinement.

Keywords: Morris water maze, spatial learning, place navigation, hippocampus, probe trial

Few behavioral tasks have shaped a field as thoroughly as the water maze has shaped the study of spatial memory. A rat is lowered into a circular pool of water made opaque with a non-toxic additive, from which the only escape is a small platform submerged a centimeter below the surface and therefore invisible. Across repeated trials the animal learns to swim more or less straight to the hidden goal from wherever it is released, which it can only do by fixing the platform's position against the stable landmarks of the room (Morris, 1981; Morris, 1984). The task is deceptively simple, and that simplicity is its power: it isolates place learning from the cues an animal might otherwise exploit, and it does so with a strongly motivating escape from water that needs no food deprivation or shock.

What the Morris Water Maze Test Is

The Morris water maze is a test of spatial learning and memory in which a rodent must learn the location of a hidden escape platform in a pool of opaque water, guided only by cues outside the pool. It is a specific paradigm within maze learning, the broad MeSH category of tasks that assess how animals learn to navigate an apparatus; what distinguishes the water maze is that the goal is hidden and the motivation is escape from water rather than food reward. Because nothing about the platform's position can be sensed directly, the animal is forced to encode where the goal lies in relation to the distal landmarks of the room — the definition of place learning (Morris, 1981).

The task's central logic is a dissociation. In the *hidden-platform* version the platform sits just below the opaque surface and can only be found by its spatial relationship to the room; in the *cued* or *visible-platform* version the platform is marked by a flag above the water and can be approached by sight alone. The two versions share every demand — swimming, motivation, vision, motor control — except the one of interest, place memory. An animal that is impaired on the hidden version but normal on the cued version has a selective deficit in spatial learning, not a general inability to perform the task (Morris et al., 1982; D'Hooge & De Deyn, 2001). This built-in control is why the maze became the field's reference assay rather than one test among many.

Place learning in the water maze is also *allocentric*: the platform is defined by its position in the environment, independent of the animal's own starting point or heading. Because release points vary from trial to trial, a strategy of “turn left and swim” cannot succeed; only a representation of the goal's location in the room — a cognitive map in O'Keefe and Nadel's sense — supports reliable escape from novel start positions (Vorhees & Williams, 2014).

The Standard Protocol

A water maze experiment has a fixed grammar. The apparatus is a circular pool, typically 1.2 to 2 meters across for rats, filled with water rendered opaque by a non-toxic additive and held at a temperature cool enough to motivate escape without causing hypothermia. A platform about 10 centimeters in diameter is submerged one to two centimeters below the surface in the middle of one quadrant. Prominent, stable visual cues — posters, shapes, equipment — are fixed on the surrounding walls; these distal cues, not anything in the pool, define the space (Vorhees & Williams, 2006).

*Acquisition* consists of repeated training trials, usually several per day over about five days. On each trial the animal is released from one of several start positions, varied across trials so that no single swim direction can be learned, and allowed to search until it finds the platform or a time limit (often 60 seconds) elapses, at which point it is guided to the platform and allowed to rest there briefly. The platform stays in the same place throughout acquisition, so performance improves as the animal learns its location: escape latency and path length fall, and swims straighten from wide, wall-hugging loops into direct approaches (Bromley-Brits et al., 2011).

The *probe trial* (or transfer test) is the critical measurement of what was learned. After acquisition the platform is removed entirely and the animal is released, usually from a novel position, for a fixed interval. An animal that has formed a precise spatial memory concentrates its search where the platform used to be, spending disproportionate time in the target quadrant and crossing the exact former platform site repeatedly. The probe trial converts the training into a memory score: it asks not how fast the animal escapes but where it believes the goal to be (Morris, 1984; Vorhees & Williams, 2006).

A well-run study also includes control conditions that rule out non-mnemonic explanations for a deficit. A visible-platform block tests that vision, swimming, and motivation are intact; measuring swim speed separately ensures that a latency difference reflects memory rather than sensorimotor slowing. Only when these are normal can a hidden-platform impairment be read as a spatial-memory deficit (D'Hooge & De Deyn, 2001).

Spatial Learning and the Hippocampus

The water maze earned its central status through a single, decisive experiment. Morris, Garrud, Rawlins, and O'Keefe showed that rats with hippocampal lesions were severely impaired at finding the hidden platform — they swam long, circuitous paths and failed to concentrate their search in the probe trial — yet were entirely normal at swimming to a visible, cued platform. The deficit was therefore specific to place navigation and could not be blamed on blindness, weakness, or a lack of motivation. This dissociation gave the hippocampus a demonstrable role in a well-defined form of memory and made the maze the standard instrument for probing that role (Morris et al., 1982).

The finding fit the cognitive-map theory of hippocampal function that O'Keefe and Nadel had advanced on the basis of place cells — hippocampal neurons that fire when an animal occupies a particular location in an environment. If the hippocampus builds and stores a map of space, then damaging it should selectively impair tasks that require navigating by that map, which is precisely what the hidden-platform maze requires and the cued version does not (Morris et al., 1982). The water maze thus became the behavioral complement to the electrophysiology of place cells, each line of evidence reinforcing the other.

Later work refined rather than overturned this picture. The maze proved sensitive not only to gross lesions but to molecular and cellular manipulations of hippocampal plasticity: blocking NMDA-receptor-dependent long-term potentiation impairs acquisition, and the maze is routinely used to read out the behavioral consequences of genetic and pharmacological interventions in the hippocampal circuit (D'Hooge & De Deyn, 2001). Adult hippocampal neurogenesis, too, leaves its signature in the maze: mice with more new neurons learn to relocate a moved platform more flexibly, implicating newborn granule cells in the fine spatial discrimination the task demands (Garthe et al., 2016).

Measuring Performance

What one records from the maze determines what one can conclude, and the field's measurement practices have grown more sophisticated as the limits of the simplest scores became clear. *Escape latency* — the time to reach the platform — is the traditional acquisition measure, but it is also the crudest: it is inflated by slow swimming, by a passive floating response to stress, and by thigmotaxis (hugging the wall), none of which are failures of spatial memory. *Path length*, the distance swum to the goal, removes the contamination by swim speed and is generally preferred as a purer index of how directly the animal navigates (Vorhees & Williams, 2014).

The probe trial yields its own family of measures. *Target-quadrant dwell time* — the proportion of the probe spent in the quadrant that formerly held the platform — is the classic spatial-memory score, read against a chance level of 25 percent when the pool is divided into four equal quadrants. *Platform crossings*, the number of times the animal passes directly over the exact former platform location, is a more spatially precise measure, and *average proximity*, the mean distance from the platform site across the probe, is more sensitive still because it credits a search centered tightly on the goal over one merely in the right quadrant (Vorhees & Williams, 2006).

Beyond these scalar scores lies the *shape* of the search itself. Animals solve the maze with qualitatively different strategies — direct spatial approaches, systematic scanning, wall-hugging thigmotaxis, circular “chaining” at a fixed radius, or undirected random swimming — and two animals with identical latencies can be using entirely different strategies. Classifying each trajectory by its geometry recovers this information, and the shift from non-spatial to spatial strategies over training is itself a measure of learning that latency alone conceals (Vorhees & Williams, 2014; Garthe et al., 2016).

Table 1

Principal Measures Taken from the Water Maze

Note. Acquisition measures track learning across training; probe measures index the retained spatial memory, in rising order of spatial precision (Vorhees & Williams, 2006; Vorhees & Williams, 2014).

Worked Example

The probe trial makes the logic of a spatial-memory score concrete. Suppose a 60-second probe is run in a pool divided into four equal quadrants, with the platform formerly in the northeast (NE, the target). A rat distributes its search time as follows: NE 27 s, NW 11 s, SE 9 s, SW 13 s. These sum to 60 s, as they must. The target-quadrant proportion is 27 / 60 = 45 percent, against a chance expectation of 25 percent (15 s). The rat spends 1.8 times the chance time in the target quadrant — clear evidence that it remembers roughly where the platform was.

The opposite quadrant (SW, diagonally across) is the natural comparison, because an animal with no spatial memory but a side bias would be caught here. The rat spends 13 s in SW versus 27 s in NE, a better-than-two-to-one preference for the correct side; a spatially naive animal would split its time roughly evenly across all four quadrants, near 15 s each. The margin between 27 s and the ~15 s null expectation is the memory signal.

Platform crossings sharpen the picture. The former platform occupies a small disc, and the pool contains three other positions equidistant from the center in the remaining quadrants. If the rat crosses the true NE site 6 times during the probe but crosses the three equivalent control sites an average of 1.3 times each, the ratio of 6 to 1.3 — roughly 4.6 to 1 — shows the search is centered not merely on the right quadrant but on the precise goal location. Quadrant dwell time establishes coarse spatial memory; platform crossings establish its precision. The two measures together, computed from the same 60-second trial, are why the probe trial, not escape latency, is treated as the definitive test of what the animal learned (Morris, 1984; Vorhees & Williams, 2006).

Discussion

The Morris water maze endures because it solved a methodological problem that had dogged the study of animal memory: how to isolate place learning from the welter of local cues, response habits, and reinforcement schedules that confound a conventional maze. By hiding the goal beneath opaque water and varying the start position, the task makes an allocentric spatial representation the only reliable route to escape, and by pairing the hidden-platform condition with a cued control it builds its own test for non-mnemonic explanations of a deficit (Morris et al., 1982; D'Hooge & De Deyn, 2001). That combination of specificity and a built-in control is rare among behavioral assays and accounts for the task's four-decade dominance.

Its very sensitivity, however, is a double edge. The water maze responds to stress, to swim ability, to thigmotaxis, and to motivation as readily as to memory, so a latency difference is only as meaningful as the controls that accompany it. The methodological literature exists largely to discipline this sensitivity: standardize the pool, the cues, the temperature, and the trial structure; measure swim speed; prefer path length and probe-trial measures over raw latency; and read a hidden-platform deficit as spatial only when the cued control is normal (Vorhees & Williams, 2006; Vorhees & Williams, 2014). Treated carelessly the maze produces artifacts; treated rigorously it remains the most informative single assay of rodent spatial cognition.

Current Directions

The most active methodological front is the automated analysis of search strategy. Rather than reducing a trial to a single latency, contemporary tools classify the geometry of the whole swim path and track how the distribution of strategies shifts across training. Machine-learning frameworks now assign each trajectory to categories such as thigmotaxis, scanning, chaining, directed search, and direct swimming, and do so with far finer granularity than earlier rule-based schemes, revealing learning dynamics that scalar scores miss entirely (Vouros et al., 2018). Open-source implementations such as Pathfinder have put this capability within reach of any laboratory, standardizing strategy classification across studies and removing the subjectivity of hand-scoring (Cooke et al., 2020).

This strategy-level view has sharpened the task's link to hippocampal biology. When the ability to switch from a coarse, non-spatial strategy to a precise spatial one is read out trial by trial, manipulations of the hippocampal circuit reveal themselves in the *kind* of search an animal adopts, not merely in how long it takes. Work on adult neurogenesis exemplifies the payoff: enriched-environment mice relocate a moved platform more flexibly, a difference that lives in their search strategies and spatial precision rather than in overall latency (Garthe et al., 2016). The direction of travel is clear — from the water maze as a stopwatch toward the water maze as a rich behavioral readout whose trajectories are analyzed with the same care once reserved for neural data.

Glossary

Acquisition.

The training phase of the water maze, in which repeated trials to a fixed hidden platform produce a progressive fall in escape latency and path length as the animal learns the goal's location.

Allocentric representation.

A map of locations defined by their positions in the environment, independent of the observer's own vantage; the kind of spatial code the hidden-platform maze requires.

Cued (visible-platform) version.

A control condition in which the platform is marked and visible, soluble by sight alone; normal performance here rules out sensory, motor, and motivational causes of a hidden-platform deficit.

Distal cues.

The stable landmarks outside the pool — posters, shapes, equipment on the room walls — against which the animal fixes the platform's location.

Escape latency.

The time taken to reach the platform; the traditional acquisition measure, but one confounded by swim speed, floating, and thigmotaxis.

Hippocampus.

The medial-temporal structure whose lesion selectively abolishes place navigation in the maze while sparing cued navigation; the neural basis of the task's spatial demand.

Maze learning.

The broad category of tasks assessing how animals learn to navigate an apparatus; the parent paradigm under which the water maze is classified.

Path length.

The total distance swum to the platform; preferred over latency as an acquisition measure because it is not inflated by slow swimming.

Place learning.

Learning the location of a goal relative to environmental landmarks, as opposed to learning a fixed response or approaching a visible cue; the capacity the water maze isolates.

Platform crossings.

The number of times, during a probe trial, the animal passes over the exact former platform site; a precise index of how tightly the search is centered on the goal.

Probe trial.

A test in which the platform is removed and search behavior is recorded for a fixed interval; the definitive measurement of spatial memory in the maze.

Search strategy.

The qualitative geometry of a swim path — spatial, scanning, chaining, thigmotactic, or random — whose classification recovers learning information that scalar scores discard.

Target quadrant.

The quarter of the pool that held the platform during acquisition; time spent here in the probe, against a 25 percent chance level, is the classic spatial-memory score.

Thigmotaxis.

The tendency to swim close to the pool wall; a non-spatial strategy, often stress-related, that inflates latency without reflecting spatial memory.

Key Researchers

Gerd Kempermann

(b. 1965). DZNE Dresden and the CRTD, TU Dresden; linked adult hippocampal neurogenesis to flexible spatial learning in the water maze, showing that newborn granule cells support the fine spatial discrimination the task demands. ORCID - Wikipedia

Richard G. M. Morris

(b. 1948). University of Edinburgh; devised the water-maze task in the early 1980s and, with Garrud, Rawlins, and O'Keefe, used it to show that place navigation depends on the hippocampus, establishing the paradigm that bears his name. Wikipedia - Google Scholar

John O'Keefe

(b. 1939). University College London; discovered hippocampal place cells and co-developed the cognitive-map theory of hippocampal function that the water maze behaviorally probes, for which he shared the 2014 Nobel Prize in Physiology or Medicine. Wikipedia

Jason S. Snyder

University of British Columbia; led the development of Pathfinder, open-source software for automated classification of water-maze search strategies, making strategy-level analysis widely reproducible. ORCID - Google Scholar

Charles V. Vorhees

Cincinnati Children's Hospital Medical Center and the University of Cincinnati; author of the standard methodological protocols for the Morris water maze, standardizing its apparatus, procedures, and analysis across laboratories. ORCID - Google Scholar

Frequently Asked Questions

What is the Morris water maze test?

It is a behavioral test of spatial learning and memory in which a rodent swims in a pool of opaque water to find a platform hidden just below the surface. Because the platform is invisible, the animal must learn its location from the stable cues around the room, making efficient escape a measure of place memory (Morris, 1981).

Who invented the Morris water maze, and when?

Richard G. M. Morris developed the task in the early 1980s, first reporting it in 1981 and detailing the standard procedure in 1984. It was quickly adopted after a 1982 study used it to show that hippocampal lesions impair place navigation (Morris, 1981; Morris, 1984; Morris et al., 1982).

Why is the water opaque?

Making the water opaque — usually with a non-toxic additive — hides the submerged platform so that it cannot be seen. This forces the animal to navigate by the room's distal cues rather than by sighting the goal, which is what isolates spatial place learning (Vorhees & Williams, 2006).

What is the probe trial?

The probe trial is a test given after training in which the platform is removed and the animal's search is recorded for a fixed interval. An animal with a good spatial memory concentrates its search where the platform used to be, so time in the target quadrant and crossings of the former platform site index what was learned (Morris, 1984).

How does the task show the hippocampus is involved?

Rats with hippocampal lesions are impaired at finding the hidden platform but swim normally to a visible, cued platform. This dissociation shows the deficit is specific to place memory rather than to vision, swimming, or motivation, implicating the hippocampus in spatial learning (Morris et al., 1982).

Why is escape latency a limited measure?

Latency is inflated by slow swimming, by passive floating under stress, and by wall-hugging thigmotaxis, none of which are memory failures. Path length, target-quadrant dwell time, platform crossings, and average proximity to the goal are more specific indices of spatial memory (Vorhees & Williams, 2014).

What are search strategies in the water maze?

Search strategies are the qualitative shapes of swim paths — direct spatial approaches, systematic scanning, circular chaining, wall-hugging thigmotaxis, or random swimming. Two animals with the same latency may use different strategies, and the shift toward spatial strategies over training is itself a measure of learning (Vouros et al., 2018).

Is the Morris water maze still used today?

Yes. It remains the standard rodent assay of spatial memory, and modern open-source tools now classify the geometry of each swim path automatically, extracting far more information from a trial than a single latency score and standardizing analysis across laboratories (Cooke et al., 2020).

References