Elevated Plus Maze: A Modern Approach to Anxiety Study

The elevated plus maze is one of the most widely used laboratory tests for measuring anxiety-related behavior in rodents, and its influence on modern neuroscience extends far beyond simply watching a rat walk around a platform. First described and validated in the mid-1980s, the test exploits a natural conflict between a rodent’s drive to explore new environments and its instinct to avoid exposed, elevated spaces. The deceptive simplicity of the apparatus, essentially a plus-shaped platform raised off the ground, belies the richness of the behavioral and physiological data it generates, and decades of refinement have turned it into a surprisingly nuanced tool for probing the biology of fear and anxiety.

How the Maze Works

The apparatus consists of four arms arranged in a cross. Two arms are enclosed by high walls, creating a sheltered corridor, while the other two arms are open with no walls at all. The entire structure sits elevated above the floor, typically around 50 centimeters high for rats and somewhat less for mice. When a rodent is placed at the center junction, it faces a choice: retreat into the safety of the walled arms, or venture onto the exposed open arms where there is nothing between it and a long drop.

Rodents reliably prefer the enclosed arms. In the original validation study, rats made significantly fewer entries into the open arms, spent less time there, and showed more anxiety-related behaviors when confined to them. Critically, confinement to the open arms also produced significantly higher plasma corticosterone concentrations compared to confinement in the closed arms, confirming that the open arms are genuinely stressful rather than merely uninteresting.¹ This avoidance of exposed, elevated space appears to be deeply wired. It does not require training, punishment, or food deprivation, which makes the test ethologically grounded in a way that many older anxiety paradigms were not.

The basic logic for interpreting results is straightforward: an increase in open arm activity, whether measured as time spent there or number of entries, reflects reduced anxiety. Drug treatments that make animals less anxious push them onto the open arms more often, while anxiogenic treatments do the opposite.²

More Than Just Open Arm Time

If all researchers tracked was how long an animal spent on the open arms, the maze would be a fairly blunt instrument. In practice, experienced observers score a constellation of subtler behaviors that reveal much more about what the animal is experiencing. Head dips, where the rodent pokes its head over the edge of an open arm and looks down, are a classic measure of exploratory boldness. Stretch-attend postures, where the animal extends its body forward from a closed arm toward an open one while keeping its hind legs planted in safety, capture a state researchers call risk assessment: the animal is actively evaluating threat without committing to full exposure.³

Pharmacological studies have confirmed that these finer-grained behaviors carry real information. When rats were given different benzodiazepine receptor compounds, the drugs not only increased conventional measures like open arm time and entries but also changed head-dipping frequency and the number of attempted open arm entries, both considered indices of risk assessment.⁴ The capacity to pick up these distinctions is part of what keeps the maze relevant: it does not just tell you whether an animal is more or less anxious, but can differentiate between drugs that act through different receptor subtypes.

Meanwhile, the physiological side of the story confirms that the maze is not merely a behavioral curiosity. Exposure to the maze for just five minutes significantly raises plasma corticosterone levels in both rats and mice compared to home-cage controls.⁵ Corticosterone is the rodent equivalent of cortisol, the primary stress hormone. The fact that a brief maze session reliably activates the stress hormone axis confirms that the behavior being measured has real physiological depth, not just a movement pattern.

The One-Trial Tolerance Problem

One of the most frustrating features of the elevated plus maze is that it essentially works only once per animal. A single prior exposure to the maze renders the rodent insensitive to the anxiety-reducing effects of benzodiazepines like chlordiazepoxide on a second test.⁶ This phenomenon, called one-trial tolerance, means researchers generally cannot retest the same animal after a drug manipulation, complicating experimental designs that require before-and-after comparisons.

The shift between a first and second trial is not just a loss of drug sensitivity. The entire behavioral profile changes. On the first trial, activity in the maze is dominated by limbic structures and reflects a generalized anxiety state. By the second trial, the prefrontal cortex and amygdala become the primary activated brain areas, and the animal’s behavior shifts from anxiety toward a more cognitively controlled fear response with more pronounced risk assessment.⁷ In essence, the animal remembers the maze and now approaches it with a different strategy. Research has confirmed that this tolerance involves learning mechanisms; blocking certain stress hormone receptors in the prefrontal cortex can prevent its acquisition.⁸

This is a genuine limitation for certain study designs. If you need to measure anxiety at multiple time points in the same animal, perhaps tracking recovery after a treatment or following the development of a condition over time, the maze’s single-use nature forces you to either use large cohorts (testing fresh animals at each time point) or look for alternative tests.

Sex, Age, and Genetic Background

The maze is sensitive enough to detect differences that have nothing to do with drugs. Female rats consistently display less anxiety-like behavior than males, spending more time on the open arms and making more open arm entries.⁹ This sex difference is not static across the lifespan, though. Up to about 60 days of age, both male and female Wistar rats show relatively high open arm exploration. The drop-off happens earlier in males, around 90 days, than in females, where it occurs closer to 120 days. Between those ages, interpreting sex differences requires caution because the two sexes are essentially at different developmental stages of their anxiety response.¹⁰

Genetic background is equally powerful. Studies comparing inbred strains of rats found significant differences across all behavioral measures in the maze, with a single factor related to approach-and-avoidance of aversive stimuli explaining much of the variation.¹¹ The same pattern holds in mice, where strains like BALB/c and C57BL/6 show distinctly different anxiety profiles, and the differences are not just in magnitude but in the structure of anxiety-related behavior itself.¹² A broader survey of 15 inbred mouse strains confirmed that strain differences appear on every measure of locomotion and anxiety the maze produces.¹³

The practical takeaway is that comparing results across labs requires knowing the strain, sex, and exact age of the animals. A finding in 90-day-old male Sprague-Dawley rats does not automatically generalize to 60-day-old female C57BL/6 mice, and this is one of the most common sources of inconsistency in the literature.

When Lab Conditions Quietly Change the Results

Even with the same strain, sex, and age of animal, seemingly minor environmental details can shift maze behavior in meaningful ways. Lighting is a well-known variable. In rats, open arm exploration increases dramatically at very low light levels (below about 1–3 lux) and then plateaus at higher illumination, behaving more like an on-off switch than a graded response.¹⁴ The finding that vision is important in triggering open-arm aversion makes intuitive sense: if you cannot see that you are on a high, exposed platform, it is less frightening.

In mice, the picture is somewhat different. A study testing C57BL/6J mice across illumination levels of 5, 100, and 800 lux found no significant effect of lighting on any measure. What did matter was the color of the closed-arm walls: mice tested in a maze with clear walls explored the open arms more than those tested with black walls, even though their stress hormone levels did not differ. Prior experience with other behavioral tests also altered maze performance.¹⁵ The wall-color finding is particularly interesting because it suggests the contrast between open and enclosed arms matters more than absolute light level for some species. If the closed arms do not look very different from the open ones, the avoidance signal weakens.

The time of day matters as well. Rats tested during their active phase (night, for nocturnal animals) show reduced open arm exploration compared to the inactive phase, and this interacts with illumination and prior test exposure in complex ways.¹⁶ None of this means the test is unreliable; it means the test is sensitive, and that sensitivity demands careful standardization.

The Brain Circuits Behind Maze Behavior

The elevated plus maze has become a key tool for mapping the neural circuits that control anxiety, partly because its behavioral readout is so clean. Recent work has focused on the prelimbic cortex, a region in the rodent prefrontal cortex, and its connections to the amygdala and hippocampus. Optogenetic suppression of prelimbic neurons projecting to the basolateral amygdala has been shown to increase time spent in the open arms, suggesting this projection normally promotes avoidance. The ventral hippocampus is another circuit node of interest, serving as a primary input to active prelimbic neurons during fear-related memory retrieval.¹⁷

Newer protocols have combined the maze with real-time optogenetic stimulation, allowing researchers to pair light delivery with specific arm entries. In one approach, the naturally preferred closed arms are selectively paired with laser stimulation of a specific brain region, creating a conflict between the animal’s safety preference and an artificially imposed aversive or rewarding signal. This lets researchers ask not just whether a circuit influences anxiety, but whether it carries a positive or negative signal that can override the maze’s built-in threat structure.¹⁸

Automated Scoring With Deep Learning

Traditionally, scoring maze behavior meant a human observer watching video and tallying each head dip, entry, and stretch-attend posture by hand. This is time-consuming, somewhat subjective, and does not scale well. Commercial tracking software offered a partial solution but has struggled with the subtler behaviors. A study comparing human raters, the open-source pose estimation tool DeepLabCut, and the commercial system EthoVision found that DeepLabCut-based tracking matched or exceeded human accuracy in the elevated plus maze, while the commercial software fell short on key measures like head dips.¹⁹

The practical difference was not just statistical. When a group of mice was given yohimbine, a drug known to increase anxiety through noradrenaline release, both human scorers and the DeepLabCut pipeline detected a significant drop in head dips, the expected anxiogenic effect. EthoVision missed it. The gap likely comes down to how each system defines head dips: humans and deep-learning approaches can recognize the posture from body-part positions, while traditional software relies on rigid spatial thresholds that miss edge cases.²⁰ This kind of improvement matters because the subtler behaviors, head dips and risk assessment, often carry more pharmacological sensitivity than crude open arm time.

The Elevated Zero Maze and Other Variations

One persistent quirk of the plus-shaped design is the center junction. Animals spend a substantial amount of time sitting in this zone, which belongs to neither the open nor the closed arms. Depending on the treatment, animals in one study spent between 13 and 30 percent of the test session in the center, effectively removing 39 to 90 seconds from the data used to calculate anxiety measures.²¹ Researchers disagree about what center time represents. Is the animal deciding? Freezing? The ambiguity creates a data-interpretation headache.

The elevated zero maze was designed to solve this. It replaces the plus shape with a circular runway, alternating between open and enclosed quadrants with no center zone. Time spent in the open sections can be expressed as a direct proportion of total test time without needing the percentage conversions that the plus maze demands. A comparison of the two designs found that the zero maze encourages greater exploration of the anxiogenic open regions and produces more consistent behavior across repeated trials, making it potentially better suited for experiments that need to assess anxiety at more than one time point.²² Despite these advantages, the plus maze remains far more commonly used, largely because of its enormous validation history and the sheer volume of existing comparison data.

Early-Life Stress as a Window Into Anxiety Development

The maze is not only used to test drugs. It has become a standard readout for studying how early experiences shape adult anxiety. Maternal separation, a widely used model in which rodent pups are removed from their mother for several hours a day during the first weeks of life, produces lasting changes in maze behavior. In one study, adult BALB/c mice exposed to maternal separation showed increased anxiety-like behavior on the maze, decreased risk assessment and exploratory behaviors, lower expression of mineralocorticoid receptors in relevant brain areas, and higher baseline corticosterone levels.²³

The effects are not always straightforward, though. A meta-analysis pooling results from dozens of maternal separation studies found that the association with increased anxiety on the maze held in rats but not in mice. Separated rats spent less time on the open arms than controls, while separated mice showed no such difference.²⁴ The species split is a useful reminder that findings from one rodent model do not automatically translate to another. Age at testing adds further complexity: one study found that maternally separated rat pups tested at three weeks showed reduced open arm entries, but those same animals tested at nine weeks showed increased entries, suggesting that the trajectory of early-life stress effects can reverse direction over development.²⁵

Beyond Rodents

The conceptual logic of the elevated plus maze, pitting a desire to explore against an aversion to exposure, has been adapted for species that never walk on elevated platforms. For fish, a submerged version of the maze uses transparent walls instead of open arms to create the exposed condition. Fish tend to avoid the transparent-walled sections in much the same way rodents avoid open arms, and the degree of avoidance serves as a proxy for anxiety level.²⁶ Zebrafish, in particular, are increasingly popular in behavioral neuroscience because of their genetic tractability and rapid development, and having a maze-based anxiety assay equivalent to the rodent version helps bridge findings across species.

At the other end of the species ladder, a mixed-reality version of the maze has been developed for humans. Participants wearing a virtual reality headset navigate a virtual elevated plus maze, and their patterns of arm avoidance mirror those seen in rodents. In a double-blind, placebo-controlled experiment, a drug that enhances GABA signaling (the same neurotransmitter system targeted by benzodiazepines in rodent studies) decreased human avoidance of virtual open arms, while a drug that blocks alpha-2-adrenergic receptors increased avoidance.²⁷ The fact that the same pharmacological manipulations shift behavior in the same direction across rats, mice, and humans lends genuine translational weight to maze-based findings. It also raises the intriguing possibility that the human version could eventually be used in clinical settings to test individual responses to anxiolytic treatments, although that application remains early-stage.

Where the Science Gets Thin

For all its strengths, the maze has blind spots. It is fundamentally a test of a single, brief emotional state: the animal’s response to a novel, mildly threatening environment over about five minutes. It does not model chronic anxiety, social anxiety, or the sustained worry that characterizes generalized anxiety disorder in humans. When a paper reports that a drug “reduces anxiety” based on elevated plus maze data, what it actually means is that the drug increased open arm exploration in rodents encountering a novel, elevated space for the first time. Whether that translates to a patient lying awake at three in the morning worrying about their finances is a much larger inferential leap than the phrase “reduces anxiety” implies.

Replicability is also an ongoing concern. The sensitivity of the test to lighting, wall color, time of day, prior test experience, strain, sex, and age means that two labs running seemingly identical protocols can get different results if any of those variables differ. Standardization efforts have improved matters, but the field still lacks a universally agreed-upon “best practice” protocol that specifies all relevant environmental parameters. This is not unique to the elevated plus maze. It runs through behavioral neuroscience as a whole. But because the maze is so widely used, its reproducibility challenges are felt across a disproportionate number of studies. The best labs treat the maze not as a turnkey anxiety detector but as one piece of a battery of tests, converging evidence from multiple paradigms before drawing strong conclusions about any drug or genetic manipulation.