KeywordsReversal Learning Giant Panda Behavioral Flexibility Cognitive Complexity Intelligent Behavior
Intelligence in animals can be defined in terms of problem-solving ability, both acquisition to solve problems and the ability to generalize what is learned to novel situations. Intelligence may also be measured in terms of behavioral flexibility, which is comprised of inhibition and innovation components and which concerns adaptability to new and changing environments.
Two main categories of inquiry derive from the central topic of nonhuman intelligence. The first category entails questions regarding species differences. Are some species of animals considered more intelligent than others? If so, which species are considered most intelligent? These questions should lead one to ponder whether it is fair to compare different species on the same tasks. The second category entails individual differences. Do individuals within a species vary in their intelligence, and how can these abilities be assessed? It is unknown to what extent nonhumans show the same variability in intelligence that humans do. Variability within a species likely depends upon the environment that it inhabits (Sol 2009). One might not expect intellectual functioning to vary within members of short-lived species whose environments are relatively constant and whose behavior is largely innate or canalized. On the other hand, one would expect variability in intellectual functioning in species that live in constantly changing environments, who face foraging challenges (Byrne 1997), and who live in complex social groups (Jolly 1966; Humphrey 1976). Humans of course live in arguably the most complex social groups and have been the most inventive in terms of extracting sustenance from the environment. They exhibit significant variability in intelligence, ranging from intellectually challenged to extremely gifted. But before delving into possible answers to these broad questions with nonhumans, it is important to define intelligence with regard to nonhumans and to consider how it might be assessed.
Intelligence is difficult to define, even in humans. Most popular tests of the nebulous construct, such as the Stanford-Binet (Roid 2003) and Wechsler Intelligence Scales (Groth-Marnat et al. 2000), focus on assessing processing speed, working memory, analogical reasoning, verbal reasoning, quantitative skills, and other basic reasoning abilities. These tests are based on the notion of a general factor, “g” that is innate, stable, and reflects an individual’s capacity to acquire and to apply knowledge (Spearman 1961). Spearman’s view also revolved around the notion that, whereas there may be different types of intelligence, they are all correlated. Other more liberal views of intelligence have been proposed such as Sternberg’s triarchic theory of intelligence that focuses on practical, creative, and analytic components that comprise what he described as “successful intelligence” (Sternberg 2012). In this view, success of the individual in real-world situations, including interpersonal and intrapersonal knowledge, are weighed just as heavily as the kinds of abstract abilities represented in most standard intelligence tests. Gardner’s theory of multiple intelligences (Gardner 2011) stressed the independence of multiple intelligence components even more heavily; this theory encompasses the view that intelligence is not a unitary construct but instead could be measured as multiple traits or abilities, such as body kinesthetic sense, musical ability, and so on, that are not necessarily correlated. That is, a person could be exceptionally skilled as a musician but not have superior verbal or quantitative reasoning skills and vice versa.
Gardner’s conception of intelligence highlights the potential for confusion between particular skills and general abilities. Should a talented musician or artist be considered intelligent in the same manner that a chemical engineer, mathematician, or neurosurgeon would be considered to be intelligent? Should distinctions be made between different types of skills with some qualifying as “intelligence” and others being considered more akin to “talents”? Both intelligence and talent seem to rely on heredity to a large extent, although talent is more likely to be honed by experience and specific training, and is always highly specified compared to the notion of g. Studies of the activation of brain areas during different types of tasks seem to support the notion of separate intelligences to some degree as they show that different areas are activated during different tasks (Hampshire et al. 2012). These recent studies emphasize the fact that the study of human intelligence is really still in its infancy despite being the target of much controversy and interest for many decades.
Researchers have not made explicit attempts to apply the same definitions to the question of animal intelligence, and, given that the topic is controversial in humans, it is clear that determining the intelligence of nonhuman organisms is not a straightforward task. However, there may be stronger agreement with a definition of intelligence that focuses on a general capacity to acquire and apply knowledge among researchers studying cognition in nonhumans. For example, working memory is commonly viewed as an essential component of animal intelligence (Carruthers 2014). Other recent approaches have focused on behavioral flexibility, which is defined as an animal’s ability to generate novel behaviors to solve problems and includes components such as inhibition along with innovation (Lefebvre et al. 2013). Whereas innovation represents the likelihood of attempting multiple novel solutions to solve problems, inhibition reflects the ability to withhold responses that may have been beneficial in the past but that no longer prove useful. Individuals may vary in the extent to which they show innovation and inhibition with each component predicting success on different aspects of problem-solving. Behavioral flexibility can be exhibited in traditional tasks, such as reversal of learning set (Roberts 1996), which requires an animal to reverse an earlier learned discrimination, learning the new task more quickly than the original task. Reversal learning, thus, clearly depends upon inhibition to a great degree.
Generalization of knowledge or concepts to novel contexts and stimuli is also a hallmark of intelligence. However, given the recent findings from neuroscience suggesting that different brain regions underlie different tasks designed to assess intelligence in humans (Hampshire et al. 2012), it seems plausible that different types of intelligence may have arisen in response to different environmental inputs for species that face unique adaptive problems. That is, the types of adaptive problems faced by, for example, insects, cetaceans, corvids, primates, and carnivores are likely to be quite distinct, leading to separate sets of adaptive skills and abilities. This fact may call into question the notion of one generalized intelligence factor that could be assessed across taxa.
What Makes An Animal Intelligent?
Along with the challenge of defining what is meant by intelligence in general, research into animal intelligence is plagued by a humancentric bias, whereby researchers often conduct “holy grail” type searches for the presence of abilities in animals that have been deemed potentially unique to humans (Vonk and Povinelli 2006). That is, humans have placed themselves atop an imaginary pedestal based on their capacity for language, tool use, imitation, culture, and theory of mind, thus setting the stage for researchers to commit careers to finding evidence for similar capacities in various other species, sometimes with little regard for whether it makes evolutionary sense for the particular species studied to exhibit the behavior in question. It is tacitly assumed that such evidence would allow for a pronouncement of “intelligence” in the species shown to exhibit it. This kind of anthropocentric focus detracts from a more ethological focus on the ability of each species to adapt to its own unique environment. Do factors that define human intelligence necessarily make sense when defining what makes an animal intelligent? For example, would an animal that spent its day calculating abstract mathematical theorems be more intelligent than the bird that was able to build a sturdy nest out of sight of predators and find food for its hungry nestlings? So, intelligence can be defined as a pattern of behavior that serves the individual in enhancing its reproductive fitness and survival or it can be defined as a suite of cognitive capacities that allow it to perform feats considered to be complex, sophisticated, and generally human-like. Different definitions may be applied within the two broad categories of inquiry identified earlier. That is, researchers may focus on the former when examining individual differences but focus on the latter when performing comparisons between species.
The study of animal intelligence has its origins in Darwin’s early fascination with the continuity of the mind across species (Darwin 1898) – a topic that was fervently explored by his successor, Romanes (1882). To what some would argue to be the detriment of unbiased study, the Aristotelian scala naturae or Great Chain of Being was at least partly responsible for the widespread idea of the hierarchy of intelligence, with humans poised atop the scale and species more closely related to humans conceived of as superior in cognitive function to those more distantly removed. Later theorists emphasized the fact that evolution should more accurately be represented as a branching bush than a ladder, with concepts such as parallelism and convergent evolution emphasizing the importance of both similarities and differences in cognitive function in animals both distantly and closely related (Novick et al. 2011). That is, distantly related species may evolve similar solutions to similar or different adaptive problems faced in their evolutionary history. Common selective pressures, such as adapting to live in large social groups (Jolly 1966; Humphrey 1976), may lead to the evolution of traits such as theory of mind and analogical reasoning in animals as diverse as corvids, cetaceans, canines, and primates.
In more recent years, researchers from various fields, such as biology, anthropology, ethology, and psychology, have all been intrigued by the question of nonhuman intelligence and have studied it from vastly distinct theoretical perspectives. Researchers studying animals in the field (e.g., Anthropologists, Primatologists, Ethologists, Biologists, Behavioral Ecologists) have tended to focus on questions of how animals adapt to and solve ecological problems, whereas those studying animals in captivity (e.g., psychologists, biologists, zoologists) have focused on questions of generalized versus compartmentalized learning.
Assessments in Nonhumans
Questions of adaptation concern the way in which an animal’s sensory and motor capacities, along with cognitive capacities such as attention, memory, and decision-making, contribute to the organism’s everyday functioning. That is, a more successful or skilled individual is better able to find food, mates, maintain and defend a territory, and raise young to reproductive age compared to a less skilled individual. Organisms will vary in the extent to which particular skills are necessary based on the particular set of environmental and social pressures that they face in their natural environment. For example, birds that nest in colonies must develop the ability to recognize individuals, including their young, whereas birds that are solitary may not need to learn to discriminate their parents from other birds and their young from other nestlings. Birds that cache or store their food over harsh winter months might be expected to exhibit superior spatial memory skills relative to noncaching bird species that live in climates where their preferred food is abundant year-round (Shettleworth 1998). Still, within these species, individuals might vary regarding the extent to which they excel at even species-typical behaviors, although it is likely that the skills most pertinent for survival are hard-wired and instinctive and would not reflect a great deal of flexibility. Thus, flexibility in behavior and the capacity to adapt to novel or changing environments seem a good candidate for something akin to intelligence in nonhumans.
However, most scientists would be unlikely to consider insects more intelligent than mammals that live in more restricted environments (e.g., polar bears, lemurs) despite the fact that insects have likely adapted to the largest number of environments in the largest quantities. There is conjecture that insects will be left to roam the earth long after humans have been wiped off the planet. If the ability of an organism to transform the environment to suit its own needs was considered the ultimate form of adaptation, then surely humans would rise to the top of the list. However, this definition might appear anthropocentric as it targets a specifically human cognitive adaptation. Whereas many species of insect exist in a multitude of habitats, only humans have proven capable of building temperature and climate controlled structures or traveling through air, ground, sea, and outer space. Furthermore, scientists might consider the Giant Panda (Ailuropoda melanoleuca) not to be “intelligent” given that it has not adapted well to varied environments and is restricted to an extremely limited diet. Given their dependence on a diet of bamboo, large territories, and a very limited reproductive window, pandas are critically endangered. It is often suggested that their appeal to humans and human concerted efforts to conserve this species is the only reason that they are not currently extinct. But cognitive tests conducted with pandas have shown cognitive skills, such as memory, categorization, and problem-solving (Perdue et al. 2009). Thus, the two different approaches to measuring intelligence are likely to lead to different conclusions regarding species differences and perhaps individual differences as well.
Psychologists and anthropologists tend to focus on questions of what it is that makes human cognition unique, which may lead to biases in the way intelligence is defined. These scientists typically define intelligence as a general capacity for acquiring information and generalizing learned rules and associations to novel events or experiences. This capacity allows an animal to behave appropriately in the absence of specific learning or associations to each independent event or stimulus. Thus transfer tests, which test the generalization of a learned rule to novel stimuli that share some feature with trained stimuli, have become the acid test of intelligent behavior for comparative psychologists. Transfer based on perceptual features is considered less cognitively sophisticated than transfer based on function or conceptual category or rule. The inference of underlying causal mechanisms that would allow the generalization of a rule or outcome is considered one of the hallmarks of intelligent behavior in humans, and researchers have sought to demonstrate this capacity in many other species with mixed interpretations of the results (Vonk and Povinelli 2006).
The application of these types of tests often necessitate that animals in the laboratory are presented with objects that are foreign to them, thus eliminating the role of associative learning and tightly canalized responding, in order to test their inferences about the function or purpose of the object. Removing the organism from contexts for which responses might be instinctive or canalized allows it to demonstrate insight and inferential reasoning. Researchers can thus disentangle intelligent behavior from hard-wired and automatic responding (Vonk and Povinelli 2010). Although a behavior can lead to success in natural environments, it may not be considered intelligent by traditional definitions if the animal need not engage in any conscious processing or decision-making to achieve the result. For example, many reflexes are highly adaptive, such as blinking when a foreign object approaches an organism’s eye, but the blink reflex is not traditionally concerned intelligent because it occurs without any conscious effort or reflection and is generally present from birth.
Furthermore, similar responses can often occur in different organisms through quite distinct underlying processes. Humans, for example, might comfort an individual who is crying because they understand and empathize with emotions of sadness, which are expressed in a crying counterpart. A feline companion might similarly appear to comfort a crying human by climbing into his/her lap and licking his/her face. But these behaviors might result from the cat’s desire for the salt in human tears or recognition of cues that the human is amendable to petting it. Tests of cognitive processes in nonhumans are challenged by the difficulty in unearthing the underlying process for behaviors that, on the surface, seem to align with behaviors that arise from complex cognitive processes in humans – the only species whose underlying thought process is easily accessible to researchers (Vonk and Povinelli 2006). Recognition of such challenges has been identified as “failure of arguments by analogy.” One simply cannot conclude that apparently similar behaviors arise from the same underlying cognitive mechanisms.
Because of the importance of underlying processes or mechanisms that support behavior, psychologists have also examined the process of learning. Animals are assessed for their capacity to learn through the use of learning sets and reversal learning. These tests determine whether organisms apply rules or strategies for learning that transcend the particular stimuli they have learned to associate or discriminate. For example, with categorization tasks, researchers aim to demonstrate that an animal can extract the features of particular stimuli and form an overarching representation of the category to which they belong. They may use two-alternative forced-choice tests in which animals are reinforced for responding to items from one category and not reinforced for responding to stimuli from another category (Holloway 1969). In addition to learning which category of items is deemed “correct” and will be rewarded when responded to, the animals might also learn the general task, that is, to extract a correct category or set of features that will result in reward and that will define all pairs of stimuli presented within the task. If researchers reverse the reward associations, the animals should learn more quickly to select the previously unrewarded stimuli in comparison to how long it took them to acquire the initial discrimination, thus demonstrating reversal learning, which reflects the application of a general rule for how to solve such tasks.
Researchers also commonly make use of matching-to-sample (MTS) tests where animals are presented with a sample stimulus and required to select the stimulus from among a set of at least two comparison stimuli that best matches the sample (Harlow 1943). Stimuli can be matched based on perceptual features, such as shape or color, or can be matched on more abstract constructs, such as sameness or differences between sets of exemplars (Bovet and Vauclair 2001; Vonk 2003). Sameness and difference can also be defined in terms of identity or similar features but also in terms of same or different relations, which can be social (Bovet and Washburn 2003) or functional in nature (Bovet and Vauclair 2001) as well. The greater the distance from reliance on perceptual features to the inference of underlying unobservable traits, the more abstract the rule, and, the more abstract the rule, the more intelligent the process is deemed to be (Vonk and Povinelli 2006).
In considering whether a species exhibits intelligent behavior or thinking, it is necessary to find only one member of the species that exhibits such a behavior, thus making it possible to test for capacities with small sample sizes in lab settings. However, then researchers must confront the risk that animals raised or housed in captivity may not be good representatives of their species, and null results cannot be taken to implicate the failure of a species to have a particular capacity (Vonk and Povinelli 2011). Laboratory studies can, on the other hand, be better candidates for exploring individual differences as individuals can easily be tracked and tested in multiple tasks. Surprisingly, few researchers have attempted to track individual differences in cognitive capacities in members of the same species to identify the possibility of a generalized intelligence factor, such as g. Vonk and Povinelli (2011) examined the findings from decades of research with a small group of chimpanzees to determine whether acquisition, transfer, retention, and overall performance were correlated across tasks in different domains (tasks that tapped into social versus physical reasoning processes) and found some evidence for individual differences as well as consistency within individuals across measures. Herrmann and colleagues conducted a large scale study of over 100 chimpanzees and over 100 human children and found evidence for a shared general intelligence that reflected physical reasoning, but a uniquely human social intelligence factor (Herrmann et al. 2010). Currently, evidence for general intelligence factors in other species is lacking. Although progress is underway, it is hampered by the difficulties in obtaining large samples of members of the same species and the time investment required to perform dozens of tests. Furthermore, the issue of comparability of tasks across species that have evolved distinct perceptual and morphological features poses a particular problem for interpreting species differences and for being certain that researchers have tapped into the appropriate tests for assessing individual differences within a species. For example, it would not be appropriate to test species like moles (family Talpidae) in a visual discrimination paradigm given their poor eyesight and then conclude that a failure to learn the discrimination reflected poor conceptual ability or that there was no variability in conceptual ability within the species. Imagine if humans were assessed for intelligence based on their ability to echolocate or to navigate using geomagnetic cues.
Brian size has often been implicated as a measure of intelligence in animals, either with or without accounting for body size. MacLean and colleagues (2014) conducted a large scale project to investigate self-control in representatives of 35 different species. They discovered a relationship between absolute brain volume and task performance. Controlling for body mass resulted in a weaker relationship between brain volume and task performance. These investigators also found that, within primates at least, dietary breadth rather than social group size was a stronger predictor of performance. Recently, researchers have examined problem-solving in a large number of carnivores and indicated a similar relationship between brain size and problem-solving success (Benson-Amram et al. 2016), this time when controlling for overall body mass. However, similar to the MacLean et al. (2014) study, there was little evidence for the importance of social group size or complexity. In future, it will be important to investigate the role of dietary factors in predicting success of carnivores in solving foraging related extraction problems and to present animals with more than a single task to assess ability (Vonk 2016). Sol and colleagues (Sol et al. 2005) have examined the role of brain size in a more naturalistic assessment of intelligence. They found that larger-brained birds adapted more successfully to novel environments compared to smaller brained birds after examining a database of over 600 introduction events. These authors were also able to show that the enhanced success of larger-brained birds could be attributed to increased innovation, which as indicated earlier is one component of behavioral flexibility.
Factors Contributing to the Evolution of Intelligence
The research summarized above highlights the factors that researchers have attended to when determining which species might be considered most intelligent. Two main hypotheses have been advanced, the social intelligence hypothesis (Jolly 1966; Humphreys 1976) and the technical intelligence hypothesis (Byrne 1997). As labeled, the social intelligence hypothesis focuses on the size and complexity of social groups in predicting cognitive ability, with larger, more complex groups predicting more advanced cognition. The technical intelligence hypothesis instead focuses on challenges faced in foraging, such as the need to find and remember the location of patchily distributed and sporadic food sources and the ability to extract food from defenses. Researchers have sometimes postulated that social complexity might predict skills in specific social domains, whereas foraging challenges may predict physical problem-solving skills, although researchers have also made the case that each factor might promote domain general intelligence. More recently Sol (2009) has revisited the cognitive buffer hypothesis, which encapsulates the notion of environmental variability in general in predicting larger brains, which, in turn, appears related to some aspects of intelligence, as described above. The theory proposes that larger brains result in an increased capacity for gathering, storing, and synthesizing information, related to the idea of domain general intelligence, and that this increased capacity may lead to a broader range of new or modified behaviors, which increase the probability of survival in complex and constantly changing environments.
More recent approaches have steered researchers away from the rigid assumptions of the past, which clung to the notion that animals advanced in cognitive complexity along a continuum leading to humans atop the imagined evolutionary ladder. Animals were assumed to exhibit greater cognitive complexity and thus intelligence, with shorter phylogenetic distance from humans. However, researchers have unearthed great behavioral diversity and cognitive complexity in animals ranging from honeybees to corvids and cetaceans. Indeed corvids and canines have surpassed great apes in some kinds of cognitive tasks, supporting the point that it is important to focus on tasks for which animals might be expected to face adaptive problems in their evolutionary history. Given the increased breadth in species of study within comparative psychology in the last decade, researchers more readily embrace the notion of convergent evolution – animals that are distant relatives and who have adapted to quite discrepant ecological niches might still evolve comparable cognitive abilities to solve problems within their own niche.
In the end, it appears necessary to consider how well animals have adapted to solve the particular kinds of adaptive problems that they will face, with a particular emphasis on flexibility to change, which will increase the odds of both individual and species survival, when determining which species and which individuals are most intelligent. As Vonk and Povinelli (2010) noted, a starting point is to observe an animal’s natural behaviors in the wild, but the laboratory may be more carefully controlled to rigorously test hypotheses about how and why specific behaviors may have evolved and how they might be altered under specific conditions. Researchers can then distinguish between behaviors that are plastic and those that are rigid and canalized. Flexible and malleable behaviors will allow an individual to adjust and adapt to an ever-changing, unpredictable environment, which is argued here to be the true mark of intelligence (Sol 2009). By combining field and laboratory approaches knowledge of what animals actually do in their natural environments can be used to further probe what they are ultimately capable of learning when taken out of those comfort zones.
- Carruthers, P. (2014). Evolution of working memory. In In the light of evolution: Volume VII: The human mental machinery (pp. 75–94). Washington, DC: National Academies Press.Google Scholar
- Darwin, C. (1898). The descent of man and selection in relation to sex (2nd ed., Vol. 1). London: John Murray.Google Scholar
- Gardner, H. (2011). The theory of multiple intelligences. In Psychology and the real world: Essays illustrating fundamental contributions to society (pp. 122–130). New York: Worth Publishers.Google Scholar
- Groth-Marnat, G., Gallagher, R. E., Hale, J. B., & Kaplan, E. (2000). The Weschler intelligence scales. In Neuropsychological assessment in clinical practice: A guide to test interpretation and integration (pp. 129–194). Hoboken: Wiley.Google Scholar
- Humphrey, N. (1976). The social function of intellect. In P. P. G. Bateson & R. A. Hinde (Eds.), Growing points in ethology (pp. 303–317). Cambridge: Cambridge University Press.Google Scholar
- MacLean, E. L., Hare, B., Nunn, C. L., Addessi, E., Amici, F., Anderson, R. C., …, & Zhao, Y. (2014). The evolution of self-control. Proceedings of the National Academy of Sciences of the United States of America, 111, E2140–E2148.Google Scholar
- Roid, G. H. (2003). Stanford-Binet intelligence scales (5th ed.). Itasca: Riverside.Google Scholar
- Romanes, G. J. (1882). Animal intelligence. London: Kegan Paul, Trench & Co.Google Scholar
- Shettleworth, S. J. (1998). Cognition, evolution and behavior. New York: Oxford University Press.Google Scholar
- Sternberg, R. J. (2012). The triarchic theory of successful intelligence. New York: Guilford Press.Google Scholar
- Vonk, J., & Povinelli, D. J. (2006). Similarity and difference in the conceptual systems of primates: The unobservability hypothesis. In E. Wasserman & T. Zentall (Eds.), Oxford handbook of comparative cognition: Experimental explorations of animal intelligence (pp. 363–387). Oxford: Oxford University Press.Google Scholar
- Vonk, J., & Povinelli, D. J. (2010). Animal intelligence. In R. J. Corsini (Ed.), Encyclopedia of psychology (4th ed.). Washington, DC: Wiley Publishers.Google Scholar
- Vonk, J., & Povinelli, D. (2011). Individual differences in long-term cognitive testing in a group of captive chimpanzees. International Journal of Comparative Psychology, 24, 137–167.Google Scholar