Cognition and intelligence

In 1871 Charles Darwin's inclusion of mind and behavior in his theory of evolution gave scientific legitimacy to the investigation of animal thinking. Since that time, animal intelligence and cognition have been of interest to psychologists, anthropologists, ethologists, biologists, and cognitive scientists. The first published treatments of animal cognition were anecdotal observations that were richly interpreted to show the complexity of animal reasoning. Those anecdotal observations were criticized for their lack of objectivity, leading to the introduction of more objective techniques such as experimental studies and more careful interpretation of results. As the field of animal cognition has progressed, emphasis on scientific rigor and objectivity has characterized research. This essay provides an overview of animal cognition as well as suggesting directions that research in the early 2000s is taking. Most of the current research and theory in animal cognition focuses on the cognitive abilities of nonhuman primates. That emphasis is reflected here. The topics chosen for this review include those that are currently dynamic and are likely to show the most growth over the next several years. This overview provides a point of entry for those interested in exploring cognitive abilities in animals.

The question of animal intelligence

Intelligence has been a difficult characteristic to define and study in humans. Various definitions and theoretical perspectives abound, addressing such issues as whether intelligence is composed of one general factor or several factors and, if several, which ones. Regardless of theoretical stance, the measurement of intelligence has been difficult and fraught with controversy. As with the concept of intelligence in humans, intelligence in nonhuman animals has been difficult to define and investigate. A definition of intelligence applicable to nonhuman animals includes aspects of learning, memory, social cognition, conceptual ability, problem-solving ability, and cognitive flexibility. Measures of intelligence must include items or tasks appropriate to the organism being studied. The determination of intelligence in animals is complicated by the lack of verbal ability in nonhumans. It is difficult to pose a question to an animal who cannot respond with a verbal answer. Investigators in animal behavior and intelligence have developed several techniques designed to understand the animal mind through directly observable behaviors. Examples of these techniques are considered below.

Brain size and the phylogenetic scale

People often characterize animals of different species as more or less intelligent, and television programs delight in questions such as "What is the most intelligent species of animal?" Scientists have looked to brain size as a way to predict intelligence across species, but absolute brain size does not work, as body sizes vary so much across animal species. The encephalization quotient (EQ) was developed as a measure of relative brain size. The EQ is a calculation based on the size of a species' brain compared to the expected size based on body size. An EQ of 1.0 indicates that the brain size is the size expected for the species body size whereas an EQ over 1.0 indicates a brain that is larger than would be expected and an EQ less than 1.0 indicates a smaller brain than expected. Humans have the largest EQ (7.0), with monkeys and apes (1.5–3.0) and dolphins (up to 4.5) also high on the EQ scale of living species. Although the EQ (as well as other measures of brain complexity) is consistent with expectations of relative cognitive complexity, it indicates nothing about the types of cognitive abilities available to animals of different species.

Historically, the study of animal cognition began with attempts to arrange species along a phylogenetic scale in order of intelligence. This approach was guided by Aristotle's proposal that organisms can be arranged along a "ladder of life" with humans at the top and other animals at different rungs, or levels, down this ladder. This concept, the scala naturae, is no longer accepted in comparative studies of animal behavior or intelligence. Rather, evolution is perceived more as a tree-like structure with individual species or groups of species branching off as they evolve adaptations that distinguish them from ancestral forms. From this perspective, animals are studied in the context of their ecological niche, the specific environment in which the organism evolved, and comparisons are made across species that share evolutionary ancestors. Each species of animal has evolved sensory capacities and a behavioral repertoire that allow success in that species' particular ecological niche. What is "intelligent" behavior for one animal

might not be so for an animal living in a different habitat. In this sense of intelligence, each species has its own type of intelligence. Investigators interested in animal intelligence have turned from global measures of intelligence and have focused research on various cognitive capacities such as basic processes like learning and memory, complex concepts like number, cognitive flexibility shown by tool use and construction, social cognition, and symbolic processing. The focus here will be on those cognitive processes.

Why study cognition in animals?

Scientific curiosity

A major reason for studying cognitive abilities in a particular species is to understand more about that species. Scientific curiosity drives many investigations of animal cognition. As we attempt to describe and understand natural phenomena, the question of the animal mind has increasingly been a subject of scientific investigation. New theoretical approaches and innovative methodologies have led to significant advances in the past 20 years in understanding how animals think. The study of animal cognition has been added to the study of animal behavior, ecology, and evolution as we continue to investigate the biology of our world.

To understand evolution

Comparing cognitive abilities across evolutionarily closely related species can provide hypotheses about how evolution occurred. Darwin's inclusion of minds in his principle of evolutionary continuity opened the area of nonhuman animal intelligence to scientific scrutiny. Because there are no fossils of behavior—only of physical structure and artifacts that imply behavioral capacities and inclinations—the relationship between structural and behavioral capacities in currently existing species is a major window into evolution. Studies of cognitive abilities across primate species provide understanding of human evolution. The search for understanding of our own species and its evolution guides much research on animal cognition. Evolutionary continuity also underlies the investigation of animal cognition in attempts to develop animal models for human phenomena. In some cases it is difficult to study a psychological process directly in humans. For example, although it is apparent that human memory relies on both verbal and nonverbal processes, it is difficult to study non-verbal memory in humans, who encode almost all information linguistically.

Conservation

Understanding the cognitive abilities of animals and how they use these capacities to solve daily problems of finding

food, evading predators, avoiding other physical dangers, and reproducing can provide important information for conservation. As habitat destruction continues to threaten the existence of many animal species, the more information available about ecology and behavior, the more informed can be plans for delaying extinction. The design and location of protected reserves relies on understanding needs of those animals being protected. Such understanding is also vital to promote the welfare of those individuals who are housed in captivity, regardless of their endangered status. Providing captive animals with cognitive enrichment by challenging their cognitive skills contributes to their psychological well-being.

Finally, as we understand more about the cognitive complexity of the animals we are attempting to preserve, the importance of ensuring their survival is apparent. These three reasons for studying animal cognition (scientific curiosity, understanding evolution, and conservation) are exemplified by the Think Tank exhibit at the Smithsonian Institution National Zoological Park. This exhibit, which opened in 1995, is dedicated to the topic of animal thinking. As the first exhibit of its kind in any zoo or public forum, Think Tank combines basic research with cognitive enrichment for captive animals while educating the public about the cognitive complexity of animals. Daily live demonstrations of data collection with orangutans (Pongo spp.) show zoo visitors how orangutans solve complex problems such as acquiring language symbols and demonstrating numerical competence. The opportunity to observe an animal using complex cognitive abilities to solve a problem not only informs visitors of the capabilities of great apes, it also serves to illustrate the importance of preserving animals with such complex minds.

Basic processes: learning and memory

Learning is generally defined as a relatively permanent change in behavior as a function of experience. Most organisms show the capacity to learn. Early studies of intelligence in animals used learning tasks to attempt to characterize species differences in capacity.

In a series of studies investigating the learning skills of rhesus monkeys (Macaca mulatta), Harry Harlow showed in 1949 that, following extensive experience with large numbers of individual problems, monkeys were able to solve a novel problem in only one trial. The problems involved making a choice between two objects that differed from each other in several physical dimensions such as color, shape, material, size, and position. A reward such as a peanut was hidden under one of the objects. The monkeys gradually learned to choose consistently the rewarded object. At that point Harlow would introduce a new problem with novel objects. Over the course of many individual problems, the monkeys took fewer trials to reach a high level of performance.

Following a few hundred such problems, the monkeys were consistently correct on the second trial of each new

problem. That is, the monkeys no longer required a period during which they learned to choose the rewarded object through trial-and-error. Rather, their response on the first trial of a new problem (whether it was correct or incorrect) informed them which object to choose on subsequent trials. This understanding of the solution to the problem based on one experience with two novel objects was called "learning set," or "learning to learn" by Harlow. It is a good example of cognitive flexibility. Learning set is still used to study aspects of learning and cognitive flexibility in humans and nonhumans. Animals of a multitude of species are capable of learning set, including cats, rats, squirrels, minks, sea lions, and several species of monkeys. The investigation of learning set in rats demonstrates the importance of considering the species-typical sensory capacities of an animal when studying cognition. Rats, who have very poor vision but excellent olfactory ability, have some difficulty with visual learning set but easily achieve high levels of performance with olfactory discriminations.

Memory provides mental continuity across time by allowing information from one point in time to be used at a later point in time. That time span can be in seconds or minutes (short-term or working memory), hours, days, or longer (long-term or reference memory). It forms the basis for learning, since without memory the influence of past experiences would not exist. Memory involves three processes: encoding, storage, and retrieval. Encoding refers to the form or code in which items or events are stored; storage involves the way that memories are stored in the brain including where, how, and how long; and retrieval refers to the act of remembering, or accessing information that was previously stored in memory. Two ways of studying retrieval are through recall and recognition measures. In recall, an individual is asked to reproduce the items or information stored at a previous time. In humans recall usually involves verbal reports. Recognition measures simply ask "Have you seen (experienced) this item before?" and involve presenting various items that were and were not in the memory set being tested. The individual is required to respond in a way that implies "yes" or "no." Because it can be nonverbal, this form of retrieval is most commonly used with nonhuman animals.

The above distinctions and methods are relevant to laboratory research with animals, and they are addressed specifically below. However, memory clearly plays a role in an animal's daily life. Food-related behaviors such as foraging or storing food for later use require memory. To what extent memory is used in foraging, and how that process is used, has been addressed in field studies of nonhuman primates. The question is whether monkeys and apes use memory to guide

their travel when foraging. For animals like gorillas (Gorilla gorilla) or leaf monkeys (Presbytis spp. and Trachypithecus spp.) whose sources of food are easily found and available in large quantities at one site ("banqueters"), remembering the location of particular food sources may not be a crucial aspect of their foraging activity. For species like orangutans and capuchin monkeys (Cebus spp.) whose food is distributed in patches that change in availability as fruits mature ("foragers"), memory of where particular trees are located as well as when they fruit may be important. Researchers have investigated whether foraging paths in social groups can be better explained by opportunistic searching or by memory-guided paths to sites that are seasonally plentiful. They have found evidence that not only do monkeys remember where plentiful sources of fruit are located, they also remember when the trees will be fruiting.

Questions surrounding memory in nonhuman animals refer to capacity, duration, and organization. These questions are addressed in laboratory experiments, some of which are designed to provide similar challenges to those provided in the animal's natural environment. A good example of this is the radial maze, developed to study memory in rats. The maze consists of a center circular area to which are attached straight alleys or runways (arms) in a manner like spokes attached to the hub of a wheel. Goal boxes at the end of each runway provide incentives. This apparatus simulates the foraging challenge presented to rats in their natural ecology. The rat is released into the center area and observed to see how it will obtain all the food. The rat's task is to run down an arm, eat the food in that goal box, return to the center area, and choose another arm to enter, repeating this until all the food has been found. Returning to an arm that was previously visited is considered an error. The most efficient solution is for the rat to run down each arm only once. To do that, the rat has to rely on memory of where it has been. Rats show effective use of memory in the task, and researchers have demonstrated that this memory is based on the formation of a spatial map of the maze. Cues from the room containing the maze are used to remember the locations of the arms that have been explored.

Research with humans has shown that humans have a large capacity for remembering items and events over long periods of time. Monkeys and orangutans have shown that they remember photographic stimuli over a delay of at least a year. Even more impressive, a chimpanzee (Pan troglodytes) showed memory for symbols learned 20 years before. Memory for concepts has been shown by squirrel monkeys (Saimiri spp.) over a five-year period, by rhesus monkeys across seven years, and by a sea lion over a ten-year period. These findings are limited only by the test intervals imposed by researchers; the limitations of specific or conceptual memory in animals have

not yet been demonstrated. It is likely that animals have very long-duration memory capacity, especially for conceptual information.

Many experimental studies with nonhuman primates have demonstrated memory phenomena similar to those shown by humans. Historically, human memory has been studied in the laboratory by providing people with lists of items to remember. When humans are provided with an ordered list of items to be remembered, they will show what psychologists term the "serial position effect." The nature of the serial position effect is that items early in the list and late in the list are remembered better than those in the middle of the list. Monkeys show this same effect with visual stimuli. The tests of monkey memory involve recognition memory while the typical procedure with humans involves recall memory. However, Lana, a chimpanzee who had had language symbol training, was able to perform a recall memory task with a list of symbols by choosing the list items from her entire vocabulary of symbols. She, too, showed a serial position effect. The serial position effect is one of the most stable phenomena in memory research with humans. To find the same phenomenon in nonhuman primates suggests that memory processes are similar across species notwithstanding the difference in language ability.

A current area of investigation in nonhuman animals is the question of episodic memory. Memory can be categorized in many different ways. The distinction between working and reference memory was discussed previously. Reference memory can also be divided into declarative (explicit, or conscious) and non-declarative (implicit, or unconscious) aspects. Declarative memory is further subdivided into episodic and semantic memory. Semantic memory refers to memory for information, in other words, generic knowledge. Episodic memory refers to memory for particular events or experiences and implies that the memory involves revisiting that event or experience. This aspect of episodic memory can be characterized as the distinction between knowing something versus recalling the specific event that provided the knowledge. Knowing that an incentive is located in a particular location without remembering the experience of seeing it hidden illustrates the distinction between semantic memory (knowledge) and episodic memory (memory for the event). Panzee, a language-trained chimpanzee, uses language symbols to indicate to an uninformed human caretaker that a particular item has been hidden at some previous time (as long as 16 hours before). Further, she will guide the human to the point where the item (that is outside of Panzee's enclosure and hence unavailable to her) is hidden. Panzee has shown the first attribute of episodic memory; the question is how to determine whether

this memory involves more than simple knowledge that a particular object is hidden outside the enclosure. That is, does Panzee's memory include "time traveling" back to the experience of seeing the object being hidden? At the present time, the answer is unclear. However, the limitations on demonstrating episodic memory in nonhuman animals are likely to be more those of procedure (how do we clearly demonstrate evidence for this capacity in nonverbal organisms?) than capacity (do animals have episodic memory?).

Number

How much do animals understand about number? Can they count? Although many animals will choose the larger of two amounts of a desirable substance, can they determine the larger of two quantities of objects? If so, what does that tell us about their understanding of number? The first animal reported to perform complex numerical tasks was Clever Hans, a horse that performed for audiences in Berlin during the early 1900s. Hans was able to answer complicated questions, including arithmetic problems involving addition, subtraction, fractions, and other arithmetic manipulations that were relatively sophisticated. When a question was posed, Hans used his right front foot to tap out the answer, and he was quite accurate. A scientific panel investigated Hans' numerical ability. They found that Hans was most accurate when his owner Mr. von Osten was present, but that he could also solve problems when Mr. von Osten was absent. This led them to the conclusion that Hans' ability was not based on fraud or tricks. However, Oskar Pfungst, a student of one of the panel members, went further in an investigation of Hans' numerical ability. He noted that Hans' accuracy depended on whether Mr. von Osten knew the answer to the question. When the answer was not known by the human questioner, Hans was inaccurate, usually tapping numbers higher than the correct answer. He also noted that there were three other individuals whose questioning of Hans was accompanied by high levels of accuracy in the horse.

With careful scrutiny of Mr. von Osten, Pfungst noted that after a question was posed to Hans, Mr. von Osten lowered his head slightly and bent forward, maintaining this position until the correct number was tapped, at which time he jerked his head upwards. The three other people for whom Hans performed well showed similar behaviors. Pfungst performed experiments in which Mr. von Osten (and the others) provided these behavioral cues at appropriate times (consistent with the correct answer) and at inappropriate times (consistent with a wrong answer). Hans' performance showed that he was responding to these cues. Pfungst suggested that Hans was sensitive to subtle cues from Mr. von Osten, beginning to tap when Mr. von Osten bent his head and continuing to tap until he detected a shift in Mr. von Osten's body posture. Hans' numerical ability was not based on an understanding of number, but rather on his reading of unintentionally provided subtle behavioral cues from his human questioner. The use of discriminative cues to control his tapping ruled out a high-level cognitive explanation for Hans' performance. Current studies of animal cognition include careful controls to eliminate the possibility of a "Clever Hans" explanation for the results obtained.

Determining what animals understand about number has several levels. The simplest level of processing numerical information is in making judgments of relative numerousness. Rats can learn to respond differentially to stimuli that vary in number, responding with a particular response, for example, to the presentation of two light flashes or tone bursts and with a different response following the presentation of four lights or tones. Rats also discriminate number of rewards and can learn sequences of reward patterns.

Monkeys and apes make relative numerousness judgments shown by their choice of the larger of two quantities of food. Chimpanzees and orangutans can make such judgments even when the two choices are themselves divided into two groups of objects, suggesting some ability to combine quantities when making the judgment. Discrimination is very accurate for quantities that differ widely from one another, such as five versus two; the task becomes more difficult as the numbers get larger and approach one another, such as five and six. If over-all mass is excluded as a possible solution by using items that vary in size (such as grapes or different candies), then a number-related cognitive ability is implicated. Most explanations of animals' success at such simple tasks include a noncounting mechanism known as subitizing, which refers to the ability to make quick perceptually based judgments of number. Subitizing is considered to be the likely way that most organisms, including human children and adults, make judgments about quantities of seven or fewer.

Understanding of ordinal relationships is a second level of numerical competence. Monkeys can order groups of items presented as abstract symbols on a video screen. They also learn to associate Arabic numerals with quantities up to nine (although this may not be the limit), and respond correctly to the ordinal positions of the Arabic numerals. Apes go further in their use of Arabic numerals as symbols by using them to demonstrate counting.

Counting implies symbolic representation of a collection of objects. The determination that an animal is counting includes three major criteria, developed initially in studying counting in human children. The one-to-one principle requires that each item is enumerated or "tagged" individually. In the stable-order principle each tag occurs in the same order (e.g., one-two-three rather than two-one-three). In the cardinal principle the final number in the count refers to the total number of items. Additional features include that any set of items can be counted, and that the order in which items are counted is arbitrary and irrelevant to determining the final count.

Chimpanzees have demonstrated all the criteria for counting. Several chimpanzees associate abstract symbols with specific quantities and act upon the symbols as they would the quantities represented by the symbols. Sheba, a chimpanzee who represents quantities with Arabic numerals, demonstrated tagging as she learned to assign Arabic numerals to collections of objects. She also spontaneously added quantities to report the sum of objects hidden across three locations.

She was able to add quantities regardless of whether objects or numerals were presented. Ai, a chimpanzee who was trained using a touch screen to respond to symbols, has shown clear understanding of ordinal positions of numerals, arranging them in order even when there are numerals missing in the list presented to her.

An interesting demonstration of the power of cognitive abstraction with numbers was shown by chimpanzees Sheba, Sarah, Kermit, Darrell, and Bobby. The task was simple but not straightforward. A chimpanzee was shown two quantities of a desirable food such as candies and was permitted to choose one of the quantities by pointing to it. The quantity that was chosen was given to another chimpanzee (or taken away), and the quantity not chosen was provided to the chimpanzee who had made the choice. This procedure works against the natural predisposition to choose the larger of two quantities, and the chimps had a great deal of difficulty with the task, even though it was clear from their behavior that they understood that they were not getting the quantity chosen. Even with much experience with the task, these chimpanzees were unable to inhibit choosing the larger quantity. When the objects presented were non-desirable objects such as pebbles, they still could not inhibit the larger choice. However, when the quantities were represented by Arabic numerals, the chimps, who were able to use Arabic numerals to represent quantities, chose the smaller number and thus obtained the larger quantity. The animals were able to use the representation of the quantities to mediate their tendency to choose the larger of two piles of objects.

This tendency to take the larger of two quantities is consistent with the chimpanzee's ecology. Chimpanzees live in social groups and compete with one another for desired food objects. The ability to judge quantities rapidly and to grab the larger of two quantities would serve them in their natural environment. The finding that orangutans Azy and Indah were able to learn relatively rapidly to inhibit the choice of the larger quantity supports the suggestion that ecology plays a role in this behavior. Orangutans are solitary in the wild and would have little need for a food-getting strategy that requires quick choice of larger quantities. Animals of both species can clearly distinguish between different quantities of food, and both can learn to inhibit the choice of the larger quantity when it is to their advantage to do so. Orangutans are able to learn to inhibit relatively quickly, but chimpanzees must rely on an additional, cognitive step in the procedure to be able to inhibit their very strong tendency to choose the larger quantity. Species difference in cognitive abilities or in the way that cognition can mediate a response can tell much about how an animal is processing information and what kinds of cognitive judgments serve the animal in its natural environment.

Tool use and construction

Tools can be defined as detached objects that are used to achieve a goal. Achievement of the goal can involve manipulation of some aspect of the environment, including another organism. The making of tools involves modifying some object in the environment for use as a tool, but not all objects used as tools by animals are constructed.

Instances of tool use and tool construction have been observed in great apes and monkeys living in captivity. All species of great ape have been observed to make and use tools in captivity. Some monkeys have been observed to use tools and occasionally to construct tools in captivity. Observations of tool use and construction by nonhuman primates in the wild have been less widespread. This difference between demonstrations of tool use in captivity and the natural environment may exist because, for the most part, evolutionary adaptations of the animal may sufficiently address challenges in the natural environment. However, environmental contingencies in captivity may be unique in encouraging tool use. In studies of tool use, researchers set up problems that are best solved using tools. In captive environments, experience with such problems coupled with extensive experience with tool-like objects provide the environmental and cognitive scaffolding that facilitate the demonstration of tool use by animals who may not so readily demonstrate such capacities in their natural environment.

However, instances of tool use and construction have been observed in great apes in the wild, particularly in the chimpanzee. The most widely known of these is the chimpanzee's use of twigs to "fish" for termites in termite mounds at the Gombe Stream Reserve (now Gombe National Park) in Tanzania, East Africa. Following the report of that first observation by Jane Goodall in 1968, many instances of tool use and tool construction have been discovered in chimpanzees and other animals in their natural environment. For termite fishing, the choice of an appropriate branch, twig, or grass blade involves judgments of length, diameter, strength, and flexibility of the tool. Any leaves remaining on a branch are removed, and the tool is dipped into holes in the termite mound. Termites attack the intruding stick, attaching themselves to it, and the chimpanzee withdraws a termite-laden stick and proceeds to eat the termites.

Additional observations of tool use include the use of leaves that have been chewed to serve as a sponge to obtain water from tree hollows, the use of tree branches in agonistic displays, various uses of sticks, leaves, or branches to obtain otherwise inaccessible food, and the use of leaves to remove foreign substances from the body. Orangutans in the wild have been observed to use leaves as sponges and as containers for foods. They use tools to aid in opening large strong-husked fruits and to protect them from spines on the outside of fruits.

An interesting complicated instance of tool use is the use of rocks as hammer and anvil to crack open nuts reported for chimpanzees in West Africa. Coula nuts or palm-oil nuts with hard shells are placed on a hard surface and cracked open by hitting them with a hand-held rock. The supporting surface (or anvil) can be a tree root or a rock obtained from the forest floor by the chimpanzee. Rocks used as hammers have a size and shape that fit the chimpanzee's hand, and rocks chosen as anvils have a flat surface. The nut is placed on the flat surface of the anvil and pounded with the hand-held hammer rock. If an anvil has an uneven base that causes it to wobble when the nut is struck, the chimpanzee finds a smaller rock to use as a wedge under the smaller end of the anvil to balance it.

Infant chimpanzees learn tool use as they observe their mothers make and use tools. Observations of mother chimpanzees engaged in explicit teaching of tool use to their infants have been reported at only one site, and on only two occasions. On one occasion, a mother took a hammer rock from her daughter who was holding it in an orientation that was not conducive to nut cracking, and slowly rotated the orientation of the rock to a position that was useful. After the mother finished cracking nuts, her daughter used the rock in the same orientation as her mother had used it. On the second occasion, a mother re-positioned a nut that her son had placed on an anvil in a position that would not have permitted its being opened. The son then used a hammer rock to open the nut. Both of these observations are provocative, but in the absence of other such observations it is not clear that the mother's intent was to teach the infant. Indeed, most learning of tool use and tool construction by young chimpanzees appears to be from observation of the mother's behavior and manipulation of objects used as tools in the context in which they are used.

Culture in nonhuman primates?

In 1953 a young Japanese monkey named Imo did something remarkable. Her troop lived on Koshima Island in Japan, and was provisioned by Japanese researchers who studied them. Provisioning involves providing additional food to sustain the population in areas of limited resources or to encourage animals to remain in a particular locale for observation. Imo's group was provided with sweet potatoes placed on the sand at the edge of the water. Imo began to use the nearby water to wash sand from the sweet potatoes. This behavior spread through the group, with younger animals adopting it first and some older animals never adopting it. Four years later, Imo introduced another novel behavior. The monkeys were provided with grains of wheat scattered on the sand that were difficult to eat because they mixed with the sand. Imo placed handfuls of this mixture of wheat grains and sand into standing pools of water. The wheat grain floated to the top and could be scooped up and eaten, free of sand. Years and generations later the monkeys of Koshima Island continue to wash sweet potatoes and to place sandy wheat grains in water. The spread of these novel behaviors through individuals in the group appeared to be an instance of social transmission of a novel behavior. This interpretation began a discussion of culture in nonhuman primate groups.

Recent analyses of behaviors shown by communities of chimpanzees and of orangutans suggest the presence of cultural variations across groups within each species living in different geographical areas. Behaviors that varied included instances of tool use, social behaviors related to grooming or

communicative signals, and food-related behaviors. For example, some chimpanzee communities crack nuts with hammer and anvil tools, but others do not, despite the availability in their environment of hard-shelled nuts and objects that could serve as tools for nutcracking. The absence of a behavior pattern in the presence of all necessary components (e.g., nuts and potential tools) rules out ecological factors to explain these differences. Similarly, genetic factors do not play a role in the variability of such behavior patterns. That is, a similar pattern of tool use may be shown by two groups who are genetically isolated from one another, or it may be present in one community but missing in another community of the same genetic background. Some form of social learning is thought to have promoted the spread and maintenance of these specific behavior patterns within certain communities.

Even in a species that is not typically group living, cultural differences are found across geographical areas. Orangutans observed in Sumatra use tools constructed from sticks to pry the seeds from Neesia fruits, a fruit with a very tough husk that also has spiny hairs protecting the seeds. Bornean orangutans do not use tools to acquire the seeds; rather, they tear a piece of the husk open to expose seeds. Only adult males can perform this latter method because of the strength required to force open the fruit. Otherwise, females and juveniles must wait until the husk opens and older, less desirable seeds are naturally available. Geographic isolation leading to a genetic basis for these differences in technique is not a sufficient explanation. At a second Sumatran site orangutans do not use tools, ruling out the genetic explanation. Neesia is available to and eaten by orangutans in both Sumatra and Borneo, eliminating an ecological explanation. Social transmission through social learning is the most likely explanation for this phenomenon, through mother-offspring transmission and/or through social encounters among animals inhabiting the same area.

Social learning can occur at several levels, from simple to complex, all based on observation of one animal by another. Most simple is social facilitation in which one animal's interest in an object elicits interest in that object from another animal. By drawing another's interest to an object that the other then explores, the first animal has influenced the behavior of the second but has not explicitly transferred information. In observational learning, the observer learns something specific about stimuli and responses from watching another animal. Imitation is the most cognitively complex form of social learning and involves the observer copying the form and intent of a novel behavior demonstrated by another animal. Imitation is distinguished from a similar form of social learning termed emulation, in which an animal performs actions similar to another, with the same intent, but without mimicking the specific actions of the model. Great apes do show imitation but that capacity may be limited to great apes.

Whether the transmission of cultural variations within a group is based on imitation or some simpler form of social learning remains unresolved. However, ongoing investigations of nonhuman primate behavior in the laboratory and in the wild will provide better understanding about the origin and transmission of novel behaviors through social groups.

Self recognition and theory of mind

Mirrors provide a novel and rich source of information about social cognition in animals. The behavior of an animal toward its mirror image suggests much about the animal's understanding of the source of that image. Many animals such as cats and dogs, when first encountering their own mirror image, behave as though they have encountered a stranger of their own species. They may show aggressive behavior such as threats, or they may attempt to play and to reach around the mirror as though attempting to find the other animal. With time, the dog or cat will ignore the mirror image and no longer attempt to engage the reflection in social interaction. For the most part, monkeys behave in a similar way to their mirror image.

Great apes, however, appear to come to understand the nature of the mirror image. That is, with experience, they behave as though they understand that it is their own body that is reflected in the mirror. The phenomenon is referred to as mirror self-recognition (MSR) and has been of interest to researchers in primate cognition for decades. Chimpanzees were the first nonhuman species to show evidence of MSR. When provided with daily exposure to a mirror outside their enclosure, individual chimpanzees initially responded to the mirror image with social behaviors suggesting that the mirror image was perceived of as an unfamiliar chimpanzee. Threat behaviors such as head bobbing, charging the mirror, and vocalizing were common. After some time, however, the social behaviors waned and the chimpanzees began to direct behavior toward their own bodies while looking into the mirror. They groomed and investigated parts of their bodies, such as the face, that were invisible to them without the use of the mirror. Using the mirror to guide their hands, these animals groomed their eyes, picked their teeth, inspected their genital areas, and also made faces while watching in the mirror.

The behaviors directed to their own bodies, termed self-directed behaviors, appeared to indicate that the animals recognized themselves in the mirror. To test this interpretation, the chimpanzees were anesthetized and an odorless red mark was placed on one eyebrow and one ear, in a location where the chimpanzee could not see the mark without the use of the mirror. When the animals awoke from the anesthesia they were presented with the mirror, and all four animals touched the mark on their brow, using the mirror to guide their fingers to the mark. The importance of this response, directing their hand to the mark on their own body rather than to the mark on the mirror image, suggests that the animals indeed recognized themselves in the mirror.

Since the initial report in 1970 of this phenomenon in chimpanzees, individuals from the other great ape species have shown MSR by demonstrating self-directed behavior or by passing the mark test, but not all individuals of these species show the capacity. There are clear individual differences based partly on age. Like humans, chimpanzees develop the ability to show mirror self-recognition. Beginning at about 24–30 months of age young chimpanzees will touch a mark on their brow using the mirror to guide the touch and the ability is generally quite evident by the age of four. Human children studied under controlled conditions do not show the capacity for mirror self-recognition until about 15 months of age at the earliest, with most achieving this developmental milestone by 24 months.

There is some evidence that dolphins are capable of mirror self-recognition, and gibbons also may have this capacity. However, other animals have not clearly shown evidence of mirror self-recognition. It may be a species difference, or it may be the case that with additional research the apparent discontinuity will be resolved. The implications for evolutionary development of self in humans are apparent, although interpretations of this phenomenon are disputed. At the most extreme, a rich interpretation of self-recognition in animals suggests an understanding of the self as an entity, perhaps similar to humans' sense of self or self-concept. However, the ability to understand the nature of a mirror image and to direct behavior back to one's own body does not necessarily imply such a rich interpretation. The distribution of this capacity and its interpretation are open questions subject to ongoing empirical investigation and theoretical debate.

The ability to recognize oneself may be related to the ability to understand another individual's knowledge state, which represents a more complex cognitive ability. This phenomenon is called "Theory of Mind" and refers to an individual's ability to understand the perspective or the "state of mind" of another. It forms the basis for such complex social strategies as intentional deception. In order to deceive another by providing incorrect information, the actor must know something about the other's perspective and expectations and what information to provide (or withhold) as deception. Many instances of deception have been reported from observations of apes and monkeys in the field and in the laboratory, and the extent to which these deceptive incidents are based on the perspective-taking capacity implied in Theory of Mind is still an open question. It is clear that animals behave in ways that suggest that they are taking into account the knowledge state of others. Whether they are or not is still an active area of research in animal cognition.

The initial description of Theory of Mind was based on the ability of Sarah, a language-trained chimpanzee, to solve problems for a human who was in some state of distress. Sarah was experienced in many features of human life. For example, she had often observed her human caretakers using keys to unlock padlocks. She had often observed humans turning a faucet to provide water through a hose. She had observed humans plugging a cord for an electric heater into an electrical outlet. Sarah was provided with videotaped instances of humans in situations whose solutions were related to the above events as well as others. For example, she saw videotape in which one of her human caretakers was apparently locked in a cage and could not escape despite attempts to open the locked door. The videotape was stopped before the problem was solved, and Sarah was provided with photographs, one of which had the solution to the problem (in this case, a key for the padlock on the cage door). Sarah consistently chose the photograph with the appropriate solution to each problem. The interpretation of Sarah's behavior was that she was able to understand the state of the human in the videotape and she was able to choose the appropriate solution to the person's problem. Additional studies with Sarah suggested she could spontaneously show deception, withholding information or providing incorrect information about the location of a food item from a human who had previously failed to share food with her. In contrast, she provided information about the location of a food object to another human who always shared food with her.

Additional studies with Sarah and other chimpanzees provided results suggestive that chimpanzees can take the perspective of another and use it to solve problems. A number of studies have been unsuccessful at providing evidence of Theory of Mind; some that have been successful have been criticized on methodological grounds. However, such studies continue. The demonstration of this capacity in apes awaits a clear methodology that will adequately address the extent to which apes can project their understanding of states of mind to others.

Social cognition

Theory of Mind is a form of social cognition, or the ability to process cognitive information presented by social partners. In group-living animals species-specific social rules guide behavior. Understanding these rules and applying them appropriately is a complicated process for individuals in the group. Recognition of familiar versus unfamiliar conspecifics, of particular age and sex classes, and of particular individuals are necessary skills for each member of the group. Even animals that spend much of their time in solitude must recognize social features of other conspecifics, including individuals who may be living in the same area. Social cognition involves not only these forms of recognition, but also understanding of species-typical social communicative signals.

In many animal species, members of the group respond quickly to alarm calls from a group member. In some cases, such response may not require much processing of information and so may be based on simple associative mechanisms. In other cases, processing of alarm calls may provide some cognitive challenge. For example, vervet monkeys (Chlorocebus aethiops) in Africa have three types of alarm calls, elicited by three different predators. Each alarm call is followed by a particular behavior by members of the group. Following a "snake" alarm call the group members stand bipedally and visually search the grass around them, presumably to locate large pythons or poisonous snakes on the ground. Looking into the air and moving to the cover of bushes follow an "eagle" alarm call. A "leopard" alarm call sends group members to trees with branches that are too fragile to support the weight of a leopard. These calls are made selectively and appropriately by adult members of the group, and receive selective and appropriate responses by group members. Infants learn the appropriate calls, sometimes making errors as they develop, for example, producing an eagle call to the sight of a harmless bird. It appears that production of the calls involves some learning, and it may be that appropriate response to the calls also is learned through observation of group members. These alarm calls are deemed cognitive rather than reflexive because production of each type of call is voluntary and the calls are referential. That is, each type of predator call refers to only one type of predator. The calls elicit different responses specific to each type of call, and production and response to each class of call show a developmental course.

Visual social signals provide information in social interactions. The simplest signals are threat or appeasement gestures. Visual social signals that guide another animal's attention to an object or event require more complex cognitive skills. For example, monkeys and apes will join another animal to investigate jointly an object of interest. Referential pointing and referential gazing are social signals that call attention to an object or event removed from the actor. Chimpanzees and orangutans can interpret pointing in humans. They also point and vocalize to draw a human's attention to a distant object or event. However, these apes do not typically use pointing to communicate with one another and their use and interpretation of pointing varies with the amount of human contact they have had.

Although dogs do not seem to respond appropriately to pointing (the usual response by the dog is to sniff the finger of the individual pointing), they do respond to gaze direction in humans and have been shown to use gaze as a cue for the location of hidden food. Similarly, chimpanzees and monkeys are able to use gaze as a cue, and monkeys will follow the direction of gaze of a conspecific, even one presented on videotape. Chimpanzees and orangutans use humans' gaze direction to locate hidden food, and as in pointing, those animals who have had extensive contact with humans are more likely to use and are more adept with gaze cues than are those with less human contact.

Language

One of the major distinctions between human and non-human animals is language. At the beginning of the animal language projects, the object was to show that an individual from a nonhuman species could acquire and use language. Apes have been the primary participants in these projects. One of the first attempts to teach language to an ape was that of Keith and Cathy Hayes, who reared Viki the chimpanzee in their home from 1947 to 1954. With much effort, Viki learned to say four words, "mama, papa, cup, and up." She had great difficulty producing these words, and they were barely intelligible. We now know that great apes lack vocal structures necessary to produce speech. However, they appear to have the cognitive ability to acquire aspects of language using symbols in communicative interactions with humans.

Washoe, a chimpanzee reared by Beatrice and Allen Gardner beginning in 1966, was taught American Sign Language (ASL). Washoe's acquisition and use of ASL were interpreted to demonstrate that a great ape could acquire a human language. The strong version of this interpretation was questioned from two perspectives. First, H. S. Terrace, who in the late 1970s trained a chimpanzee named Nim Chimpsky to learn ASL, questioned the referential nature of a chimpanzee's use of ASL. Terrace suggested that the primary basis for Nim's "linguistic" ability was not linguistic or referential in the way that human language is used to refer to objects, concepts, and experiences, but was more likely based on associative learning and imitation in goal-oriented situations. That is, Nim appeared to reply to questions posed by his teachers in a manner that was more imitative of the teacher's gestures than suggestive that he was generating linguistic utterances of his own that referred to objects, concepts, and experiences in his environment.

A second related question addressed syntax, or grammar. Chimpanzees who learned ASL (and there were more than Washoe and Nim involved in such studies) did not follow strict grammatical rules of word order required in human linguistic utterances, whether spoken or signed. Rather, the chimpanzees would repeat words or phrases, often varying the order of words as the utterances were repeated. Such behavior was consistent with Terrace's suggestion that the chimpanzees were not using signs to communicate. Beginning in 1973, Duane Rumbaugh trained a chimpanzee named Lana to produce sentences through the use of a computer-operated keyboard containing abstract symbols that represented verbs, nouns, adjectives, and adverbs. These symbols, called lexigrams, were composed of abstract geometric shapes combined to form unique configurations, each of which referred to a particular word or concept. Rumbaugh developed a language he called "Yerkish" named for the Yerkes Regional Primate Research Center where the Lana project began. Lana's symbols followed rules of grammar and the results of this project showed that Lana and other chimpanzees in the project could learn rules of syntax and use abstract visual symbols in a communicative manner. Chimpanzees Sherman and Austin later showed the ability to use this computer-based system to communicate information to each other.

Although for some investigators of animal cognition the question of ape language is still at issue, the ape language projects have expanded in content and have provided important discoveries about ape cognitive abilities. In such a project, Chantek, an orangutan reared by Lynn Miles beginning in 1977, learned ASL. In this project Chantek's acquisition of symbolic communication was studied in the context of over-all cognitive development. This project is unique in its breadth, providing understanding of the development of symbolic communication as one feature of a suite of cognitive abilities.

In 1971, David Premack first reported results of a project in which he trained chimpanzee Sarah to communicate using arbitrary abstract symbols. These symbols were colored plastic shapes. Sarah learned concepts such as "same" and "different," as well as nouns, verbs, and adjectives. That these

symbols were representational to Sarah was demonstrated when she was provided with the symbol for apple, and asked to describe its physical features. Although the symbol for apple was a blue triangle, Sarah described the object presented to her as "red" and "round," referring to attributes of the object rather than of the symbol. She also showed the ability to reason analogically through her understanding of "same" and "different." The primary contribution of this language project was not Sarah's linguistic abilities, but how Premack used Sarah's representational capacity to show the breadth of cognitive flexibility and conceptual understanding available to a chimpanzee provided with Sarah's rich cognitive environment.

In a similar fashion, two current ape language projects use the animals' symbolic ability to uncover cognitive representational abilities that would be difficult to access without this means of symbolic communication. In 1995 Robert Shumaker began to work with orangutans Azy and Indah, who are learning an abstract symbolic communication system presented as lexigrams on a touch-sensitive video screen. When questions are posed, the orangutans indicate the appropriate symbol from an array of symbols by touching it. Shumaker is using the animals' symbolic abilities as a window into other related cognitive processes such as number comprehension. In a project begun in 1978 by Kiyoko Murofushi, Toshio Asano, and Tetsuro Matsuzawa, chimpanzee Ai continues to demonstrate her sophisticated cognitive abilities using a touch-sensitive video screen and a vocabulary of lexigrams and Arabic numerals. For example, Ai labels objects and their characteristics such as color and number; she clearly understands numbers conceptually, and counts using Arabic numerals from zero to nine; and she "spells" by constructing lexigrams from their components. Ai's infant Ayumu, born in 2000, is learning to use the touch-screen system, providing insight into the development of symbolic cognitive skills as he interacts with his mother and observes her using the system.

The symbolic ability of great apes in the context of a language-learning setting was most strongly demonstrated by Sue Savage-Rumbaugh who in 1979 extended the Lana project to another great ape species, the bonobo (Pan paniscus). Kanzi, a young bonobo who lived with his mother during her training with Rumbaugh's computer-based language, surprised researchers when he showed clear understanding of the task and of particular symbols simply as a result of having been passively exposed to the symbols while his mother was learning the task. Following this discovery, Savage-Rumbaugh focused on Kanzi's ability and motivation to use abstract communicative symbols. Kanzi not only uses the abstract symbolic language system, he also has a demonstrated understanding of spoken language, and Savage-Rumbaugh has reported that he appears to attempt to communicate vocally by imitating acoustic properties of human speech. Her interpretation of Kanzi's behavior is that bonobos appear to have not only the ability to acquire abstract human-derived symbols, but that they may have additional communicative skills that can provide insight into the evolution of human language.

Koko the gorilla has been using ASL under the tutelage of Francine Patterson since 1976. She is probably the most publicly recognizable language-trained ape. Indeed, Patterson has extended her project into conservation efforts in the United States by promoting Koko to the public. Further, Patterson translated a children's book about Koko into French to distribute in French-speaking Africa as a way to educate children about the cognitive and emotional capacities of gorillas and the importance of preserving them in the wild.

The importance of the ape language projects has not been in the demonstration of a human capacity in great apes, but rather in exposing the complex symbolic skills available to animals in a rich interactive environment. The cognitive skills shown by the nonhuman participants in these projects extend beyond the specific symbolic skills trained in individual projects. They have opened a window into the minds of animals that enriches our understanding of the animal mind, while also providing new theoretical perspectives on the evolution of human cognition and language.

Enculturation of apes

Much of the information we have about cognition in great apes has come from research projects involving intensive study of one or two apes, usually chimpanzees. In most of these research projects the apes have had extensive interaction with humans during their early development. These interactions not only take the form of explicit cognitive and behavioral tests, but they also include teaching and guidance. The apes acquire sophistication with human artifacts including computers and other electronic or mechanical objects.

The environment for these animals is complex, and not typical of the environment in which apes evolved. Although some of the ecological challenges met by animals living in the natural environment are missing, the early experience of these animals is enriched and challenged in different ways. Apes who experience this intensive experience with humans are called "enculturated" apes, those who at some level have been exposed to and integrated into certain human social/cultural experiences. Because they have not been exposed to their natural environment, some researchers question the value of the conclusions based on their cognitive abilities. Clearly, these animals have had enriched early experiences, including direct teaching of skills by human caretakers, and are not representative of chimpanzees developing with their mothers in the wild environment. This challenge has been met by the response that although the cognitive tasks provided to enculturated animals differ from those in the natural environment, they challenge the cognitive skills required for survival in the wild.

Further, the possible enhancement of cognitive skills provided by the enriched environment shows the extents and limits of cognitive abilities under circumstances conducive to high levels of performance. The results of these studies tell us what cognitive skills great apes are capable of; the extensive behavioral studies of great apes in their natural environment show us how they use these cognitive abilities to solve daily social and environmental challenges. Although an enculturated ape has had a different early environment, he or she continues to be an ape and to show cognitive skills available to an ape. Integration of data from field and laboratory continues to provide the richest understanding of great ape cognition.

Resources

Books

Bekoff, M., C. Allen, and G. M. Burghardt, eds. The Cognitive Animal. Cambridge, MA: MIT Press, 2002.

Candland, D. K. Feral Children and Clever Animals. New York: Oxford University Press, 1993.

Hauser, M. D. Wild Minds. New York: Henry Holt and Company, 2000.

Matsuzawa, T., ed. Primate Origins of Human Cognition and Behavior. Tokyo: Springer-Verlag, 2001.

Mitchell, R. W., N. S. Thompson, and H. L. Miles, eds. Anthropomorphism, Anecdotes and Animals. Albany, NY: State University of New York Press, 1997.

Parker, S. T., and M. L. McKinney. Origins of Intelligence. Baltimore: The Johns Hopkins University Press, 1999.

Parker, S. T., R. W. Mitchell, and H. L. Miles, eds. The Mentalities of Gorillas and Orangutans in Comparative Perspective. Cambridge, UK: Cambridge University Press, 1999.

Pearce, J. M. Animal Learning and Cognition. 2nd ed. Hove, UK: Psychology Press, 1997.

Premack, D., and A. J. Premack. Original Intelligence: The Architecture of the Human Mind. New York: McGraw-Hill/Contemporary Books, 2002.

Roberts, W. A. Principles of Animal Cognition. New York: McGraw-Hill, 1998.

Shettleworth, S. J. Cognition, Evolution, and Behavior. New York: Oxford University Press, 1998.

Tomasello, M., and J. Call. Primate Cognition. New York: Oxford University Press, 1997.

Wynne, C. D. L. Animal Cognition. Basingstoke, UK: Palgrave, 2001.

Karyl B. Swartz, PhD