Skip to main content

History of Mazes


History brought to life. MazeEngineers dove deep into the literature in this online museum.


Brought back to life, view, discover, and even order the maze. Request here at MazeEngineers

Cincinnati Water Maze

Cincinnati Water Maze is commonly used tool for the studying of egocentric and spatial navigation as well as learning, and memory of a subject.

Double Alternation Spatial Maze

This is a well-constructed device that will meet all your scientific needs.

Maze Engineer’s Modern Version of Small’s Hampton Court Maze

Hampton Court Maze

The Hampton Court Maze is a reliable tool for observing the spatial learning and behaviour in rodents.

Sensory Control Of The White Rat In The Maze Habit

Dennis 1929: The Sensory Control of the White Rat in the Maze Habit

Tridimensional Maze

The tridimensional maze is like the bidimensional maze in that the correct pathway leads constantly into new portions both of the maze and the environment.

If you were to conjure an image of a researcher studying behavior in a laboratory, one of the most reliable answers would surely be the clichéd scenario of a scientist donning a white lab coat and scribbling notes on a clipboard while observing a mouse or a rat wind its way through a maze,

history of behavioral testing with mazes using rats

on a hunt for a cheesy reward. Save for a few aesthetic diversions, that cliché is not all that far removed from the reality of behavioral testing. Mazes are one of the most commonly employed methods in behavioral research and for good reason. Current mazes have been designed to investigate many facets of rodent behavior. They can be used to study innate psychological states, such as anxiety, and critical mental processes like spatial navigation, learning, and memory. The maze has become such an essential part of behavioral research that it can be easy to take its existence for granted. Like all the best scientific tools, the maze was not born overnight. In fact, the mazes we know today are the product of over a century of development in which they have grown both in complexity and versatility. In this series of posts, we will explore the science and the minds that developed rodent mazes into this valuable behavioral research tool.

The use of rodent mazes began in the laboratory of Dr. Edmund Sanford at Clark University in the 1890s. One of his graduate students, Linus Kline, was interested in studying what was then known as zoological psychology, akin to what we know today as comparative psychology. Animals had long been used as models

for the study of human physiology, and Kline was inspired to use them as behavioral models by Darwin’s work in the mid-1800s. Once in Sanford’s lab, Kline began work on a thesis titled “The Migratory Impulse vs. the Love of Home,” which would focus on the migratory impulses of animals.

Linus Kline’s early work for his doctoral thesis utilized different animals for experimental and observational research, including tadpoles, chicks, and even human children. He eventually moved towards studying rats, partly because of their known behavioral patterns, and in part, because they are “small, cheap, [and] easily fed and cared for…”[1]; all are essential aspects of a behavioral model in modern labs. The first studies he completed in rats were not mazes in the traditional sense, but rather puzzles. He designed a behavioral apparatus intended for the observation of a natural rat behavior: gnawing and digging. Kline wished to create conditions in which a rat would learn to associate these natural behaviors with sating hunger. To do so, he placed a hungry rat beside the apparatus, which contained a food reward – which was, in fact, cheese. The rat could gnaw its way into the box, or dig its way to the reward, but it had to learn to do so. Once placed by the apparatus, the experiment proceeded much in the same way as it does today; Kline fastidiously took notes as he observed the rat until, in his words, “it grew quite monotonous watching them”[1] (a sentiment with which graduate students over a century later still likely identify).

Figure 1: Sketches in a letter from Linus Kline[1].

Kline would later decide to employ mazes to study the home-finding behavior of rats in collaboration with another of Sanford’s graduate students, Willard Small. While discussing the capacity of rats to burrow underground when going back to their nests, Sanford suggested using the pattern of the Hampton Court maze – a popular tourist attraction at the time – to create a navigation test (i.e. maze) for rats to solve in the laboratory. Small was interested not in the natural home-finding or food-seeking behavior of rats, but rather in their learning processes. Though their goals were different, a maze was poised to suit both of their research needs. Small’s experiments with mazes would go on to be the first ever published study that utilized a maze as we know it today.

Small wished to study the natural mental processes of the rat. Like Kline, he used hungry rats motivated by a food reward, but he was not interested in the association of food reward with a natural impulse like digging or gnawing. He chose to construct a maze because reaching the food reward meant learning the solution to a spatial puzzle. It entailed a series of deliberate decisions that would lead the subject to its reward.

While researchers today are accustomed to taming laboratory rats, Small initially trained wild rats in his maze. Subsequent experiments with the maze would be performed with the more familiar tame white rat because the wild subjects were tricky to handle. The nature of much of Small’s study is comparative and observational. How did an individual performance change with experience? How did domestication influence the performance of the animals? Did males and females exhibit the same behavior in the maze? These were the first iterations of questions that behavioral researchers still ask when they perform maze experiments today, albeit as a smaller component of larger questions.

A sizeable portion of Small’s publication of this first rodent maze is dedicated to his interpretation of the role that the rat’s different senses played in solving the maze. One group of his rats went naturally blind, allowing him to draw conclusions about the role of sight in maze-solving; these blind rats were, in fact, able to solve the maze just as the normal rats did. His experiments were not conducive to

Figure 2: The first published rat maze designed by Willard Small[2].

rigorous study of each of the rat’s senses and their relative contributions to maze-solving, however. While he was able to make some thoughtful interpretations to this end, he suggests that his explanations of what the different senses may or may not contribute only serve to highlight the complexity of such behavioral testing, calling them a “warning against naively accepting simplest explanations.”[2]

It is critical that scientists understand the relative contributions of the senses if they are going to attempt to draw conclusions about higher cognitive processes such as learning or memory. If a rat learns that it simply must travel a certain distance and then turn, is it truly learning to solve the maze, or is it forming a simpler association based upon a proprioceptive clue? One of the first scientists to address this question using a maze was Dr. Walter S. Hunter. He performed many experiments in the first half of the 20th century which together cast a lot of doubt on the proprioceptive theory of maze-solving. This is the concept that a rat can solve a maze simply by piecing together a series of movements – for example, walk this distance then turn right, walk that distance then turn left. Other sensory stimuli may play a role in solving the maze, but this theory posits that proprioception is a principal component of the strategy.

Figure 3: A version of Hunter’s double alternation tridimensional maze[3].

Dr. Hunter published his tridimensional double alternation maze in the 1920s in an effort to directly address questions of how sensory processes factored into maze-solving. His setup was unique. The mazes were constructed so that each turn was exactly the same distance apart. Every angle in the maze was a right angle. The floors and walls were carefully built and maintained to eliminate tactile differences that could signal location. The experimenter was always in the same place. There were no drafts that could signal direction. There were no shadows. In some versions of his maze, the food reward was not even kept in the maze, but rather away from it to avoid olfactory interference. Any noises that could possibly be heard, like the sound of cars on the street, were carefully noted in case they could be used as a clue. The environment was heavily controlled. The strategy for his maze was one of double alternation, meaning two left turns would then be followed by two right turns, then two left turns, and so on. He found that they could solve this in two dimensions, so he progressed to a tridimensional maze, in which the exit was located directly above the entrance.

The results of Dr. Hunter’s study produced substantial variation between subjects. His mazes were very difficult for some rats to solve, but the bottom line was that even in these heavily controlled sensory conditions, his mazes were solvable.

He strongly believed that even the sense of proprioception could not have been used to make this maze easier to solve. The same distance had to be traveled between every turn, and how could that motor sequence produce a left turn sometimes and a right turn other times on its own? Perhaps other maze conditions in previous studies had allowed the rats to make these simple associations and then solve the maze. In these conditions, the rats had to work much harder. At its core, Dr. Hunter showed that it was possible to construct mazes that forced the utility of higher mental processes than sensory association.

Maze Engineers’ Modern Version of Linus Kline’s Mouse Box

Maze Engineer’s Modern Version of Small’s Hampton Court Maze

Small, Hunter, Kline, and Sanford are among the very first researchers to pioneer the use of rodent mazes in behavioral research. In the next section, we’ll explore how that groundbreaking work kept pace with the scientific community’s growing computational power through the contributions of scientists like Dr. Claude Shannon – the father of information theory.

A Mechanical Maze Solver – Mid 1900s

Today’s scientists have surely noticed a push towards computational modeling. In neuroscience, for example, computational approaches are combined with new and existing biological or behavioral data to form algorithms and other models that can detail complex cognitive processes. In the mid-1900s, after mazes had been used in behavioral research for

Claude Shannon and mouse in history of mazes

Figure 4: Claude Shannon and his mouse, Theseus [4].

several decades, the famous scientist Dr. Claude Shannon brought a computational approach to maze learning. He even brought a rodent along with him, in the form of a little mechanical mouse named Theseus.

Shannon is known today as the “father of information theory.” He was a highly regarded mathematician and engineer in the mid-twentieth century and worked as a cryptographer during World War II, when scientists like him were called upon to decipher coded enemy messages. Shannon developed Theseus in the 1950s while working at Bell Telephone Laboratories. Theseus was designed to solve mazes using a system of switching relays like the ones found in old dial telephone systems. The system would allow him to learn any combination of moves to solve a maze. Once learned, you could place Theseus in any arbitrary point of the maze, and he would know how to get to the finish line. We have an advanced version of Theseus known as MINA.

The maze itself was similar to what we know as the Hebb-Williams maze; the outer boundaries of the maze are fixed, but the inner walls can be moved and rearranged into any number of combinations. If the maze had changed from the last time Theseus ran it, he would turn in place to explore the new square before proceeding. On the next run, he would get through that maze quickly and with no errors. These exploratory turns can be seen in traces that were captured from above. The tracking, in this case, is not unlike

the tracking we use today. Instead of a motion sensing camera above the maze, Theseus carried a small lamp and was photographed at long exposure to capture the trace.

Shannon originally developed Theseus in an effort to further understand how the technology that drove those old telephones carried information, so that, ultimately they could make it better and more efficient. The mouse in a maze is truly an apt metaphor to connect this endeavor to behavioral science. When Dr. Willard Small first published the rodent maze, he wanted to understand the mental processes of the rat during natural behaviors. Today’s neuroscientists join behavioral research with computational research to achieve exactly these goals within the context of the brain. How does the brain, with its finite number of cells, synapses, and neurotransmitters, possess the ability to produce and transmit such an incredibly wide array of information? Theseus used his set of switching relays to make a decision in Shannon’s maze. We are still using mazes to understand how a real, living animal processes information and makes decisions.

Figure 5: A photographed trace of Theseus exploring a maze [4].

Maze Engineer’s Advanced Version of Theseus – MINA

Claude Shannon recognized the potential of merging new technology with established protocols to investigate new questions. In the next section, we’ll take a look at how mazes have continued to evolve in the modern era.

An Enduring Technique – Mazes in Modern Times

The use of animal models for studying human physiology enjoys a long history – over thousands of years[5]. Relatively speaking, mazes are still rather new, having only been adopted just over a century ago. However, the scientific world moves rapidly when it comes to methodologies. The cutting-edge technology one day could be considered outdated ten or fifteen years down the line. Yet the maze is still commonly employed by behavioral scientists to this day, due in no small part to its flexibility to be adapted to modern work.

The employment of the rodent maze as a scientific method has two foundational levels of flexibility that have allowed it to keep pace with the times. The first level has to do with the maze itself. A great many varieties of maze exist today for use in rodent research, each specifically designed to probe distinct behavioral traits. The capacity to design a maze to suit specific scientific inquiries has been appreciated almost since the dawn of its creation; Dr. Walter S. Hunter exemplified this concept as he adapted his simple and double alternation mazes into three dimensions.

Today’s scientists still have the option of designing their own mazes for their studies, though many opt to leverage the wide variety of tried-and-true mazes that have been published over the years. Given a behavioral trait, there’s a good chance that a maze has already been designed to investigate it. If it’s spatial navigation or other forms of learning and memory you wish to study, you may opt for a radial arm maze, a Morris water maze, or even a Hebb-Williams style maze, like Dr. Claude Shannon, among other options. For more straightforward decision-making studies, a T-maze or Y-maze may be the right fit for the experiment. Even more effective processes have mazes designed to investigate their biological and psychological underpinnings. Anxiety, for example, can be quantified more or less in rodents using elevated plus or zero mazes.

The second key aspect of rodent mazes that have rendered it an enduring technique are the rodents themselves. The mouse particularly has become a fixture, not only in behavioral science but in biological research more broadly. Due to the impressive technological advancements that have been made with the mouse, the mouse is an applicable model organism to an incredible variety of studies. Part of this model animal’s success is a product of its extensive characterization. One could fill an encyclopedia with the amount of knowledge we have collected about mouse behavior and physiology as well as its precise genome. Thanks to decades of inbreeding, scientists know exactly what they’re working with when they have a lab mouse, genetically speaking.

It is hard to adequately describe how essential the full characterization of the lab mouse genome is to expanding the boundaries of feasible research. This knowledge has enabled us to map behaviors onto specific genes that have then been relevant to human physiology. For example, this command of the mouse genome allowed scientists to map disrupted circadian rhythms onto a gene we now call Clock. Just this year, the value of such work investigating the mechanisms controlling circadian rhythms was recognized with the Nobel Prize in Physiology or Medicine. Our fluency in the mouse genome has also allowed scientists to manipulate the genome to create superior models of human disease. Modern tools have enabled the creation of “humanized” mouse models – in other words, transgenic mice that have been engineered to express human genes. Such mice are regularly characterized using a battery of behavioral tests, often including mazes, to assess the effect of these human genes in behavior and disease progression.

Scientists have also developed tools that allow them to manipulate mouse brain functions in ways that can be temporally controlled. Some of these tools, such as optogenetics, allow a scientist to promote or inhibit the firing of a neuron on a near-immediate timescale by shining a laser onto a part of the brain that has been engineered to express genes that will respond to that particular wavelength of light. Other tools, such as Designer Receptors Exclusively Activated by Designer Drugs (aka DREADDs), enable manipulation of brain activity on a timescale more akin to drug activity. Mice engineered with DREADDs have unique receptors in specific areas of their brains that work like a lock and key; certain cell types are given this distinct lock (the DREADD), and those locks can only be opened by one key (a specific drug administered by the scientist). When the lock is “opened,” the cell responds appropriately.

Animals engineered for optogenetic or DREADD experiments often undergo behavioral characterization to see what the impact of manipulating those brain cells will have. For example, a scientist could train a mouse engineered for an optogenetic test to run a maze until it has mastered it. Then, the scientist could place the mouse at the start but this time turn on the laser in an area crucial for spatial navigation, like the hippocampus. Suddenly, the mouse may forget where it must go, highlighting the importance of this brain region for spatial navigation. Likewise, a mouse engineered for DREADD experiments could be given the DREADD-activating drug before a maze run to assess the impact of that brain region between trials.

Scientific tools are constantly developing and changing. The cutting-edge technology of one day can be rendered obsolete by progress just a decade later. The methods that stand the test of time are the ones that can address fundamental questions in versatile ways and evolve with time. Over the last century, scientists have shown that mazes possess exactly those qualities. While the bells and whistles appended to the rodent maze have changed with time to meet growing scientific demands, the fundamental aspects of the maze have remained relevant and valuable with the passage of time.


  1. Miles, W.R. (1930). On the History of Research with Rats and Mazes: A Collection of Notes. The Journal of General Psychology 3(2): 324-337.
  2. Small, W.S. (1901). Experimental Study of the Mental Processes of the Rat II. The American Journal of Psychology 12(2): 206-239.
  3. Hunter, W.S. (1929). The Sensory Control of the Maze Habit in the White Rat. The Pedagogical Seminary and Journal of Genetic Psychology 36(4): 505-537.
  4. Cybernetic Zoo (2010). 1952 – “Theseus” Maze-Solving Mouse – Claude Shannon (American).
  5. Ericsson, AC, Crim, MJ, & Franklin, CL. (2013). A Brief History of Animal Modeling. Mo Med 110(3): 201-205.
Close Menu