Animal Cognition

Bradley R Sturz, Kent Bodily, Michelle Hernández, Kelly Schmidtke, Jeffrey S Katz. 21st Century Psychology: A Reference Handbook. Editor: Stephen F Davis & William Buskist. Volume 1. Thousand Oaks, CA: Sage Publications, 2008.

Animal cognition refers to the mechanisms by which nonhuman animals process information and is primarily concerned with understanding how animals learn, store, remember, and respond to information from the environment. Although similar in many respects to the present-day fields of animal learning, behavioral ecology, and ethology, animal cognition has emerged as a distinct area of study by adopting the information-processing perspective from human cognitive psychology. The information-processing perspective assumes that animals form and manipulate internal representations about their environment. As such processes afford flexibility and adaptability in behavior necessary for survival in an ever-changing environment, the information-processing perspective has proved useful in explaining and predicting numerous behavioral phenomena in animals.

Historically, animal cognition finds its roots in the American tradition of animal learning (e.g., Edward Lee Thorndike) and the European tradition of ethology (e.g., Konrad Lorenz, Niko Tinbergen). Each tradition differs in approach and method of investigation. Whereas ethologists often stress adaptive specializations and focus their studies in the context of an animal’s natural environment, those studying animal learning often stress general processes and focus their studies in the context of a laboratory. Despite the rich historical roots in these traditions, animal cognition did not emerge as a distinct field of study until shortly after the cognitive revolution of the 1960s. It was then that the information-processing perspective was infused into the fields of animal learning and ethology. Since that time, animal cognition continues to mirror its human counterpart in focus and theory while maintaining an interest in learning and behavior. Over the years, these traditions have been integrated into a contemporary synthetic approach (Kamil, 1988; Shettleworth, 1998) that attempts to elucidate both specialized and general cognitive processes in the laboratory and field.


Today, information-processing and evolutionary perspectives (influence of selective pressures in shaping cognitive mechanisms) serve as theoretical frameworks for research on animal cognition. From these frameworks, two main theoretical approaches for understanding psychological phenomena in animals have been developed: modularity and general process. One of many current challenges facing today’s scientists is discovering procedures that can distinguish between these two approaches (but see Bitterman, 1965, 1975; Kamil, 1988).


The modularity account of psychological phenomena proposes that mental processes are domain-specific mechanisms called modules (e.g., Fodor, 1983; Shettleworth, 1998). Overall, modularity proposes that animals utilize a number of these psychological mechanisms (modules) for the purposes of solving specific environmental problems. Each module processes a specific type of information. For example, just as an anatomical mechanism such as the eye is highly adept at processing a specific type of information (light), psychological mechanisms (modules) are adept at processing specific types of information. Given that species adapted to flourish within an ecological niche, their unique cognitive modules also may have evolved to solve environmental problems specific to that niche. An implication of a modularity account is that different species may perform qualitatively differently when engaging in analogous tasks or problems because of these unique cognitive mechanisms.

General Process

The general-process account of psychological phenomena proposes that mental processes, instead of being species-unique, are qualitatively similar across various species (e.g., Bitterman, 1960; Papini, 2002). Given that species developed specialized adaptations to flourish within unique ecological niches, similar cognitive mechanisms also may have evolved either through homology or homoplasy to solve common environmental problems. An implication of a general-process account is that evo-lutionarily divergent animals may perform qualitatively similarly when performing in analogous tasks or problems because of these similar cognitive mechanisms.

General Methods

Although investigations of animal cognition also occur in an animal’s natural environment, the present chapter focuses primarily on laboratory research. To study animal cognition in the laboratory, researchers precisely control experimental conditions in a host of procedures and tasks, and although apparatuses vary widely, the most common is the operant chamber. The operant chamber is a completely enclosed rectangular space that allows for the presentation of stimuli and recording of responses. Stimuli are shown on a computer monitor or behind translucent buttons located in the front panel of the chamber. A touch screen surrounding the monitor (e.g., like that of an ATM machine) or the buttons themselves detect responses to the stimuli. From these responses, researchers determine measurements such as accuracy and reaction time. Such precise control of an animal’s environment allows researchers to rule out or control factors (e.g., past experience) that often confound experimental results from the natural environment.

A unifying theme of animal cognition research is that experimenters measure animal behavior during the learning of some task. Usually, researchers implement specific performance criteria for the task. The performance criteria ensure that an acceptable level of learning occurs. Upon reaching the performance criteria, researchers often introduce novel situations. By measuring accuracy and reaction time during these novel situations, researchers can infer what the animal is learning and how it is solving the task. In short, these novel tests under steady-state conditions reveal how the animal encoded and processed information.

Areas of Interest

Numerous areas of research are open to those interested in animal cognition, including (but not limited to) attention, perception, memory, spatial cognition, timing, counting, concept learning, problem solving, consciousness, emotion, foraging, theory of mind, communication, language, and tool use (for reviews see Roberts, 1998; Shettleworth, 1998; Wasserman & Zentall, 2006). Below we introduce several of these research areas and briefly discuss common methodologies and general results unique to each.


The current study of animal memory closely mimics that found in human cognition and places emphasis on how previous experiences are represented, stored, and remembered. Thus, like research on human memory, research on nonhuman memory also involves the development of tasks that assess the dynamic processes of encoding, retention, and retrieval. Encoding refers to the transformation of sensory stimuli (e.g., visual, auditory, olfactory, gustatory, and tactile) into mental representations. Retention refers to the storage of these representations, and retrieval refers to the reactivation or later access of these representations. Researchers often subdivide memory into various memory systems (or cognitive modules), including separate systems for working memory and long-term (reference) memory. Reference memory consists of declarative and procedural memory, and procedural memory consists of semantic and episodic memory. Procedural memory refers to memory for cognitive and behavioral skills, whereas declarative memory refers to memory for environmental information. Semantic memory refers to memory for general information about the environment, whereas episodic memory refers to memory for specific events (i.e., the what, where, and when of an event). Evidence supports the existence of all of these memory categories for animals, albeit the data are controversial in some cases (e.g., episodic memory).

Methods and Results

Common methodologies used to assess memory include delayed-matching-to-sample (DMTS) tasks, symbolic matching tasks, list-memory tasks, radial-arm mazes, and cache recovery tasks. To test working memory, researchers often utilize a DMTS task. In a DMTS task, a trial begins with the presentation of a sample (e.g., a blue circle). To ensure the subject encodes the sample, a response to the sample stimulus (e.g., a touch) occurs, after which the sample disappears. After an interval of time elapses (known as the retention interval), two stimuli are shown (known as comparison stimuli). One of the comparison stimuli is the same as the sample and the other is different. A correct choice response, a response to the comparison stimulus identical to the sample, generally results in a food reward. Experimenters often manipulate the length of the retention interval and the number of stimuli that serve as the sample. By then measuring the influence of these variables on animals’ performance, researchers are able to determine both the duration and capacity of working memory.

Much like DMTS tasks, symbolic matching tasks involve the presentation of a sample and a choice among comparisons. The main difference between the two tasks is that in the symbolic task the sample and the correct comparison are not identical. Instead, the researcher establishes the samples and their correct comparisons arbitrarily. For example, an animal may be shown a red square as the sample, then a horizontal line and a vertical line as comparisons. The animal may then receive the reward only after selecting the horizontal line. Because the sample and comparisons are not identical, perceptual similarities such as hue or brightness cannot be the basis of a comparison choice; thus, correct responding on symbolic tasks suggests that animals memorize (reference memory) the individual samples and their correct comparisons.

List-memory tasks also begin with the presentation of a sample stimulus. After the observing response, the sample disappears. Instead of the presentation of the comparisons, as in a DMTS task, another sample is shown that also requires an observing response. After a number of sequential sample item presentations (i.e., the list), researchers introduce a retention interval followed by a single probe item. The probe item can be either identical to one of the list items or different. The animal then responds according to whether the probe item was presented in the list of sample items. For example, a pigeon may peck a right button if the probe item was presented in the list or a left button if the probe item was not presented in the list. As with DMTS and symbolic tasks, the list-memory task is useful in determining both memory duration and capacity. However, the list-memory task also has the distinct advantage of revealing changes in performance as a function of list-item position. For example, list-memory tasks have revealed primacy (better memory for items present early in the list) and recency effects (better memory for items present late in the list) with both pigeons and monkeys, which parallel results from human participants in similar tasks (Wright, 2006).

The radial arm maze is an apparatus that consists of a central starting position with an even number of arms emanating from its center. All of these arms are accessible from the central starting position. Before a trial, the experimenter places food at the end of the arms (and, at times, in either a fixed or random pattern). An animal (usually a rat) is put into the start position and then searches for the food rewards. When searching for the food, rats remember which arms they previously visited, as they rarely make a repeat visit to an arm (Olton & Samuelson, 1976). This result occurs even if the animal is taken from the apparatus for a fixed duration (retention interval) before it continues searching. Although such performance by rats may be the result of fixed patterns of searching (procedural memory), it also may be the result of memory for locations visited within the maze (declarative memory).

Researchers studying episodic memory in animals often use cache recovery tasks. A typical cache recovery task involves allowing an animal to store food at specific locations and then measuring search location accuracy on later retrieval of these items. Clayton and Dickinson (1998) initially found that scrub jays (a food-storing bird) can learn that the decay timeline (i.e., how long it takes a food item to become inedible) for wax worms is more rapid (< 124 hours) than that for peanuts. These birds then cached both wax worms and peanuts in visuospatially distinct locations and later recovered these items after 4 or 124 hours. After delays of 4 hours between caching and recovery, scrub jays searched more often at the worm cache locations, whereas after delays of 124 hours they searched more often at the peanut cache locations. Such results suggest that the scrub jays encoded location (where), content (what), and duration (when) of their cache sites.

Spatial Cognition

The study of spatial cognition focuses on how animals acquire, process, and store spatial information (Roberts, 1998). Such information is critical to the survival of mobile organisms because it allows for the learning of locations and permits navigation between locations. Although researchers routinely investigate several aspects of spatial cognition (e.g., orientation, spatial memory), the bulk of spatial cognition research emphasizes mechanisms of spatial coding of position. Two separate reference systems for spatial coding of position are egocentric and allocentric (for a review see Newcombe, 2002; however, see Wang & Spelke, 2002 for a view based on only an egocentric system). Researchers define these systems with respect to how spatial locations are referenced—either to the self (termed egocentric) or to the environment (termed allocentric). Based on these reference systems, evidence supports the use of several processes for locating objects in the environment such as path integration, beacon homing, and landmark-based navigation.

In the absence of landmarks, many animals are able to return to a location from which they start (e.g., their nest). Such a phenomenon, called dead reckoning, occurs by path integration—a process in which an estimation of position is updated via speed of movement, direction of movement, and elapsed time of movement (Etienne et al., 1998; Etienne & Jeffery, 2004). However, if a landmark is available, and located at or near a goal, some animals can plot a course directly to the environmental stimulus. This beacon homing process serves as a simple and effective way to locate places in the environment. If, however, a landmark is some distance from the goal, animals are able to rely on landmark-based navigation. Landmark-based navigation is the process of determining a position and orientation by using objects in the environment with known positions (Gallistel, 1990).

Methods and Results

Researchers interested in spatial cognition use several methodologies. The most common apparatuses include water mazes and open fields, and the common component to almost all spatial cognition tasks involves animals searching for a goal location. For example, in a standard Morris water maze task (Morris, 1981), researchers place animals (usually rats or mice) into an enclosed circular pool containing murky water. A small visible platform in the pool serves as the goal location. The amount of time to reach the platform is the measure of learning, and, not surprisingly, animals quickly learn to swim to the platform to escape the water. Across trials the platform progressively becomes invisible (by submerging it). Various cues placed around the walls enclosing the pool or on the inside walls of the pool itself serve as the only consistent source of information as to the location of the platform. Under these circumstances, animals continue to locate the invisible platform. Even in the absence of the platform, the animal spends the majority of search time near the absent platform’s location. Locating the invisible platform and searching in the appropriate area for the absent platform provide evidence that the animal relies on the cues surrounding the pool to locate the goal. By then moving or removing these cues, the experimenter is able to determine which specific cue(s) the animal uses to locate the goal. Overall, experiments using the water maze show that animals rely on both local cues (stimuli close to the goal location) and global cues (stimuli at some distance to the goal location) but show preferential responding to closer and larger stimuli.

An open field also is routinely used in the study of spatial cognition and consists of a search space covered with a material (known as substrate) such as sand or sawdust in which an animal can search. Initially, researchers place food above this substrate at a specific distance and direction from one or more landmarks. Progressively, on subsequent trials, the researchers bury the food and the animal must search for the hidden food. Animals are adept at locating the food and continue to search at the correct location even on trials in which food is absent. Searching at the goal location despite the absence of food suggests that animals do not rely on visual or olfactory cues to locate the food. Instead, animals utilize information from the landmarks. Much like the water maze, open field tasks reveal which cues (landmarks) animals use to locate the goal but, importantly, also are ideal for determining how animals use these landmarks. For example, pigeons trained to find a goal in the presence of a single landmark will adjust their search behavior accordingly following shifts of the landmark to novel locations (Cheng, 1988, 1989). Such shifts suggest that animals encode both the distance and direction of the goal location from the landmark (Cheng & Spetch, 1998).

Concept Learning

The study of concept learning focuses on how animals learn to categorize stimuli and make judgments about relationships between stimuli. Concept learning research falls into three broad groupings: perceptual concepts, associative concepts, and abstract concepts. Perceptual concept learning involves sorting (or categorizing) stimuli (e.g., pictures of birds, flowers, people, trees) by stimulus similarity into appropriate categories (Herrnstein, Loveland, & Cable 1976; Huber & Aust, 2006). Associative concept learning involves sorting stimuli by a common response or outcome regardless of perceptual similarity into appropriate categories (e.g., Urcuioli, 2006; Vaughan, 1988). Abstract concept learning involves judging a relationship between stimuli by a rule (e.g., sameness, difference, addition, subtraction). Researchers consider a rule abstract if performance with novel stimuli during transfer tests is equivalent to performance with training stimuli. Although there is little dispute among theorists concerning the ability of various animals to learn associative and perceptual concepts, controversy remains concerning whether animals can learn abstract concepts because of alternative accounts (Katz, Wright, & Bodily, 2007). Nonetheless, evidence for all three types of concept learning occurs in species as diverse as bees, capuchin monkeys, dolphins, parrots, pigeons, rhesus monkeys, and sea lions. Such research also reveals that the number of items present during training (also known as set size) can be critical to concept learning (Shettleworth, 1998).

Methods and Results

Researchers use a variety of methods to test for concept learning, and two of the more common include sorting tasks and the same/different (S/D) task. Using a sorting task, Bhatt et al. (1988) allowed pigeons to see a single visual stimulus from one of four categories: car, cat, chair, or flower. The pigeons then selected one of four simultaneously available buttons that each corresponded to one of the four categories. After learning to categorize the initial training images, the researchers introduced novel images from each category. Pigeons correctly categorized the novel images better than chance, suggesting that they learned the perceptual concepts of car, cat, chair, and flower. Additional studies found that increasing the number of training images (set size) improved transfer performance (e.g., Wasserman & Bhatt, 1992). Presumably, correct performance increased because novel pictures became more similar to the training stimuli with the increase in the number of training stimuli.

In the S/D task, researchers present an animal with two items within a single trial. Sometimes the pair of items are the same (e.g., a baseball and a baseball) and other times they are different (e.g., an apple and a grape). Depending on trial type (i.e., same or different), the animal receives reward for making the appropriate response. For example, an animal might first see two pictures on a computer monitor. Next, a green square and blue diamond are made available simultaneously to the left and right of the pictures. A response to the green square results in reward when the pictures are the same, and a response to the blue diamond results in reward when the pictures are different. To test whether animals simply memorize the training item pairs, researchers introduce novel pairs of items (items different from those during training). If the animal discriminates novel pairs as well as the training pairs, it is taken as evidence for abstract concept learning. When the training set (e.g., 8 items) consists of a small number of training pairs (e.g., 64 pairs: 8 identical and 56 different) animals do not transfer to novel items, but when the training set becomes progressively larger (16, 32, 64, 128, 256) the level of transfer increases to the point where it is equivalent to training performance (i.e., full concept learning; Wright & Katz, 2006).

Although set size has similar effects on perceptual and abstract concept learning, the mechanisms that underlie these two types of concepts are different. Specifically, researchers claim the set size effect occurs because of the similarity between training and transfer stimuli for perceptual concept learning and by the formation of a rule between stimuli for abstract concept learning.

Timing and Counting

Timing refers to a process by which an animal tracks and marks how “much” of a time interval or duration passes. Counting refers to a process by which an animal tracks and marks how “many” of a quantity occurs. Although timing and counting may be separate mechanisms, both involve differentiating an amount, and this functional similarity suggests that these processes may result from a common mechanism.


There are at least two types of timing: circadian and interval. Circadian timing relies on a single pacemaker—a biological oscillator that pulses within the body and marks 24-hour durations. Circadian timing is involved in cyclical patterns of behavior such as sleep/wake cycles. These cycles are thought to result from the synchronization of the pacemaker with a reliable external cue such as the sun. Interval timing also involves a pacemaker to determine the amount of time that elapses between discrete events.

Methods and Results

Researchers use a variety of methodologies to assess timing. For example, circadian timing studies might present a signal, such as a light, at an irregular time. If a light is presented at the beginning of the night to a nocturnal animal, such as a hamster, it will rest longer into the following day. It is as if the presentation of light offset the pacemaker and shifted the entire cycle forward in time.

Investigations of interval timing may use conditional discrimination tasks. For example, rats in an operant chamber may first learn one response (e.g., depression of a lever) after the presentation of a two-second tone and another, different response (e.g., depression of a different lever) after the presentation of an eight-second tone. To assess the accuracy of an animal’s ability to time, investigators present new durations of the tone. During these critical test trials, animals often respond to the lever associated with the two-second tone for shorter durations and the lever associated with the eight-second tone for longer durations. This reliable, differential responding suggests animals track the interval duration.

Another method to investigate interval timing uses a fixed-interval schedule, where a stimulus is present (e.g., a light or a tone) and food reward follows the first response emitted after a specific amount of time. For instance, a pigeon may receive reward for the first peck that occurs five seconds after presentation of a tone. Not surprisingly, responding increases in number as the interval progresses, suggesting that the animal is tracking the amount of time elapsed since the presentation of the stimulus. Another procedure, the peak procedure, modifies the fixed-interval procedure by inserting a trial with a longer duration that occurs in the absence of food reward. Here, animals tend to respond the most at the point in time at which food should occur, and responding decreases after that point. Such response patterns suggest that animals track the amount of time elapsed as they are able to determine when food should occur.


Recall that counting is a process by which an animal tracks and marks quantities. Gelman and Gallistel (1978) proposed five criteria necessary to claim that an animal is capable of counting. The one-to-one principle states that each item in an array should receive a distinct tag. For example, in English the tags are one, two, and three. The stable-order principle asserts that the tags must have a reliable order. One always precedes two and two always precedes three. The cardinal principle states that the last tag represents the total number of items in the array—in this case, three. The abstraction principle requires that counting not occur by rote memorization—any array with three items must be responded to as three (i.e., transfer). Finally, the order-irrelevance principle states that as long as each item has a single tag and tags follow a reliable order, which object is given which specific tag is irrelevant. Such strict criteria force researchers to develop precise procedures for training and testing counting behaviors.

Methods and Results

Methodologies to study counting can be similar to conditional discrimination tasks used to test timing with the exception that experiments use numerosities instead of intervals. For example, Brannon and Terrace (2000) used a task with rhesus monkeys that required correctly making a sequence of responses. Specifically, the researchers presented the monkeys with four colored boxes that each contained one to four items. The size of the items within the box and the location of the boxes varied to ensure that the monkeys attended to the number of items within each box and not the amount of space taken up by the items or the location of the boxes. The researchers required the monkeys to touch each box once (the one-to-one principle) in a reliable ascending order (the stable-order principle) and end on the highest quantity (the cardinal principle). When novel items were presented in the boxes, the monkeys performed above-chance accuracy, indicating the learning was not purely due to memorization (the abstraction principle).

Problem Solving

Problem solving refers to a process in which animals overcome obstacles in order to obtain an outcome (Köhler, 1927). Problem-solving research seeks to delineate tasks that animals are capable of solving and attempts to explain how the animals solve these tasks. The design of experimental problems demands great care and creativity from researchers and often results in novel and interesting methods for testing the problem-solving abilities of animals.

Problem solving involves several components: representation, learning, and choice. These components serve as the focus for much of the problem-solving research (Lovett, 2002). In short, representation refers to information encoded internally; learning refers to changes in behavior resulting from experience, and choice refers to the selection of specific behaviors or objects among a multitude of possibilities.

Methods and Results

Methodologies for studying problem solving are numerous and diverse. As the development and selection of problem-solving tasks vary widely due to researcher preference and species under study, we present a select few of the more well-known, problem-solving tasks in lieu of a list of common procedures.

Wolfgang Köhler, when studying chimpanzees, devised a now-famous problem commonly known as the “box and banana” problem. In this problem, he suspended a banana in the air so as to be unreachable and then scattered various objects including wooden boxes about the immediate area. The chimpanzees initially attempted to obtain the banana by unsuccessfully jumping for it. However, one chimpanzee, Sultan, began to look back and forth from the banana to the nearby boxes. All of a sudden, Sultan grabbed a box, put it under the banana, climbed atop the box, and grabbed the banana. As if through a flash of insight, problem solving was sudden and complete. However, despite observing the entire process, the other chimpanzees did not execute the task. Other chimps, although engaged in behaviors associated with components of the task, failed to link the separate skills successfully to obtain the banana. For example, one chimp found a box that was not near the banana, climbed on top of it, and stared at the distant banana.

In other studies, Köhler continued to present animals with the task of acquiring food out of their reach. As with the “box and banana” problem, objects such as sticks or boxes were within the animal’s immediate reach. Much like with the “box and banana” problem, the animals were unsuccessful in their initial attempts at reaching, stretching, grabbing, or jumping for the food. However, a few animals obtained the food by utilizing the available objects. Importantly, manipulation of the location of the object in the chimpanzees’ visual field increased their success rates. Specifically, objects that were at locations too far out of the visual field resulted in lower success rates. Köhler suggested that the changes in the environment influenced the animal’s representation of the problem and thus its ability to arrive at an effective solution. In other words, representational changes (changes to internal codes of information) resulted in higher or lower success rates. This line of research provided the foundation for research on problem representation and its importance in the problem-solving process.

Thorndike’s (1898, 1911, 1932) famous “puzzle box” studies involved placing a hungry cat in a box with food outside of its reach. The only way for the cat to obtain the food was to trip a lever, pull a string, or push a pole inside the box. These various behaviors served to open a door, allowing the cat access to the food. Thorndike noted that initial attempts to escape the box were sporadic and haphazard. In time, the cat performed the proper operation and escaped the box. Thorndike also noted that as the number of trials within the apparatus increased, the time to escape decreased; the animals’ behaviors became more direct at activating the mechanism responsible for opening the door. From these observations, Thorndike argued that problem solving was a gradual process. In his opinion, this “trial-and-error” learning was responsible for the production of behaviors that eventually resulted in solutions. Perhaps, most important, this line of research led Thorndike to develop one of the most basic principles of learning, the Law of Effect, which states that responses that result in satisfying consequences are strengthened.

In other studies investigating the learning component of problem solving, Robert Epstein and colleagues (1984, 1985) carefully controlled the prior experience of pigeons before testing them on an analog of the “box and banana” problem. Pigeons learned (a) to push a box toward a green dot at random locations, (b) to step up onto a box under a small facsimile of a banana hanging from the ceiling of the operant chamber, and (c) to not fly or jump at the banana. After learning these separate behaviors, a pigeon would enter the chamber with the banana hanging overhead and the box’s location off to one side. However, the green dot would be absent. Stepping up on the box did not allow the pigeons to reach the banana (as the box was not under the banana). All three pigeons solved the problem within about a minute! Perhaps most important, pigeons with slight changes in training (e.g., pigeons that learned to push the box but not toward the green dot; pigeons that learned to peck the “banana” but not step up onto the box) failed to solve the problem. Epstein and his colleagues concluded that an ability to solve a problem depends largely upon specific prior learning experiences.

A series of laboratory studies with wild-captured New Caledonian crows nicely demonstrates the choice component in problem solving (Chappell & Kacelnik, 2002; Weir, Chappell, & Kacelnik, 2002). In these experiments, a small bucket with a wire handle (like a small tin paint can) containing meat was lowered into a plastic tube. The crows were unable to reach the meat or the wire handle with their beaks because the tube was too deep. However, the researchers placed two pieces of garden wire next to the tube: One piece of wire was straight and the other was bent into a hook. Impressively, numerous crows selected the hook-shaped wire, extended the hook into the tube, and retrieved the meat-filled bucket.

Tool Use

Although problem solving and tool use involve many of the same components, modification, in contrast to simple selection, is necessary for a behavior to be considered tool use. For example, in other experiments with the crows mentioned in the problem-solving section, some individuals actually bent a straight piece of wire into a hook and proceeded to retrieve the bucket from the tube (Kenward, Weir, Rutz, & Kacelnik, 2005). As another example, some chimpanzees in the wild remove a small branch from a tree, break off all smaller branches from it, and then insert the thin flexible branch into a termite nest. Chimpanzees eat termites, and not surprisingly, the termites readily cling to the branch. Thus, the chimpanzees need only to remove the branch from the termite nest to obtain the food.

In both of the cases just mentioned, the animal uses a modified object to achieve a result. However, many examples of tool use exist, especially in the wild, that are not as clear-cut. For example, chimpanzees use broken tree branches like a hammer to crack nuts. Crows drop nuts onto rocks (and pavement) to crack them. Sea otters swim on their backs and carry a flat stone on their stomachs. The stone serves as an anvil against which to beat clams and expose the meat. Some troops of capuchin monkeys beat clams against tree trunks. In all these examples, animals clearly use objects or structures in their environment to solve a problem, but the behaviors are not examples of tool use because the objects used as “tools” are not modified. These and other examples continue to challenge the way we think and talk about tool use in animals. It is important to remember that regardless of how we as humans define tool use, animals will continue to use objects in their environment, some of which they will modify, to obtain food or solve other problems.

Methods and Results

As with problem solving, methodologies used for studying tool use are quite diverse. However, one of the more common experimental tasks is the “trap tube task.” The trap tube itself is a hollow, foot-long, cylindrical piece of transparent plastic about an inch in diameter with a small reservoir (the trap) located directly in the middle. The tube is often mounted horizontally on a foundation, and researchers place a food reward inside the tube next to the reservoir. In order to obtain the food, the animal (often a primate) must insert an available stick into the appropriate end of the tube or else the stick will push the food into the reservoir. In this task, the use of the stick itself to obtain the food falls into the gray area between problem solving and tool use but certainly does not entail modification. However, when given a bundle of sticks tied in the shape of an H, many animals will remove the unnecessary sticks and produce one stick to insert into the tube (Visalberghi, Fragaszy, & Savage-Rumbaugh, 1995).


Although the study of animal cognition is important in its own right, it has many important applications. For example, research illuminating mechanisms underlying animal cognition are especially relevant to human education and mental health. Specifically, the study of animal cognition often results in comparative models of cognition. If normal cognitive functioning can be well understood in simpler cognitive systems, it is often possible to adapt this knowledge to more complex cognitive systems. This fundamental knowledge about normal cognitive functioning may permit identification of cognitive deficits and, more important, the potential sources of these deficits. As a result, fundamental knowledge about the nature of cognition also has potential for advancements in technological and methodological innovations for improving education, diagnosis, and treatment of learning disorders, mental disorders, brain-based injuries, and degenerative diseases. It may also provide researchers in other areas of scientific study with invaluable opportunities to explore associated neurological mechanisms and efficacy of pharmaceuticals.


Emanating from the American tradition of animal learning and the European tradition of ethology, the study of animal cognition is a field focusing primarily on understanding how animals learn, store, and respond to environmental information. Animal cognition has emerged as a distinct field of study due to the infusion of the information-processing perspective from human cognitive psychology into the fields of animal learning and ethology, coupled with its focus on general and specialized psychological processes in the field and laboratory. This approach results in specialized methods and tasks for studying such cognitive phenomena as memory, spatial cognition, timing, counting, concept learning, problem solving, and tool use. The information gleaned from investigations into animal cognition has important implications for gaining fundamental knowledge about the nature of cognition and its potential for advancements in technological and methodological innovations, such as improving education and the diagnosis and treatment of learning disorders, mental disorders, brain-based injuries, and degenerative diseases.