Within the animal kingdom, complex brains and high intelligence have evolved several to many times independently, e.g. among ecdysozoans in some groups of insects (e.g. blattoid, dipteran, hymenopteran taxa), among lophotrochozoans in octopodid molluscs, among vertebrates in teleosts (e.g. cichlids), corvid and psittacid birds, and cetaceans, elephants and primates. High levels of intelligence are invariantly bound to multimodal centres such as the mushroom bodies in insects, the vertical lobe in octopodids, the pallium in birds and the cerebral cortex in primates, all of which contain highly ordered associative neuronal networks. The driving forces for high intelligence may vary among the mentioned taxa, e.g. needs for spatial learning and foraging strategies in insects and cephalopods, for social learning in cichlids, instrumental learning and spatial orientation in birds and social as well as instrumental learning in primates.
Neuroscientists hold that higher cognitive functions, also called ‘intelligence’ or ‘mind’ (as defined in the next paragraph), are bound to brain properties, and that degrees in cognitive functions and intelligence correlate well with degrees of brain complexity. For a long time, the common view even among biologists was that the joint evolution of brains and minds started with diffuse nerve nets and very simple behaviours like those found in acoelans and culminated in a straightforward fashion in the human brain as basis for the superior mental abilities that make humans ‘unique’ . However, despite their indisputable complexity, the human brain and mind are the result of just one out of many lines of evolution. Complex brains are supposed to have evolved at least several times independently, often in distantly related taxa, although presumably from an ancestral tripartite brain , and with them high intelligence. A detailed presentation of the content of this article can be found in my book ‘The long evolution of brains and minds' .
2. What is intelligence?
In humans, intelligence is commonly defined as the sum of mental capacities such as abstract thinking, understanding, communication, problem-solving, reasoning, learning and memory formation and action planning. Usually, human intelligence is measured by intelligence tests and expressed in intelligence quotient values expressing different contents (e.g. visual–spatial, verbal, numerical). Evidently, such a definition and measurement of intelligence cannot be applied directly to non-human animals. A number of comparative and evolutionary psychologists and cognitive ecologists converge on the view that, across all species studied, mental or behavioural flexibility or the ability of an organism to solve problems occurring in its natural and social environment are good measures of intelligence culminating in the appearance of novel solutions not part of the animal's normal repertoire [3–5]. This includes forms of associative learning, innovation rate as well as abilities requiring abstract thinking, focused attention, concept formation and insight.
3. Brain complexity
With respect to bilaterian animals, the term ‘brain’ is used for a condensation of nerve cells around the oesophagus giving rise to nerve cords . There is no universally accepted definition of brain complexity. However, most concepts depart from the idea that—besides absolute brain size or (uncorrected or corrected) brain size relative to body size (cf. )—the number of neuronal and non-neuronal cells inside the brain, the number and pattern of short- and long-term connections, the number of anatomically and functionally different cell types and the degree of compartmentalization (major parts of the brain, nuclei, modules, layers, etc.) are basic neuroanatomical criteria for brain complexity. At a functional level, one can take into account the number and diversity of neuroactive substances (neurotransmitters, neuropeptides, neurohormones), differences in axonal conduction velocity, modes of synaptic and extrasynaptic transmission, etc. However, because of the lack of a both formalized and empirically founded theory of brain complexity, we have to restrict ourselves to relatively informal comparisons regarding brain complexity, especially when comparing distantly related taxa.
4. How many times did complex brains and ‘high intelligence’ evolve?
The origin of the first brains and their structures is disputed. Cnidarians as the assumed sister group of bilaterian animals have no brains, although local concentrations of nerve cells can be found. Among the Acoelomorpha, the Acoela, but not the Nemertodermatida and Xenoturbella, are assumed to possess a brain-like rostral concentration of nerve cells plus subepidermal nerve chords. Brains are believed to be present in the last common ancestor of protostomes (Lochotrophozoa, Ecdysozoa) and deuterostomes (Ambulacraria, Chordata), which may already have had a tripartite organization . This would imply that the ambulacrarians (echinoderms, hemichordates) have lost a brain. Alternatively, brains have evolved independently in protostomes and chordates from diffuse nerve nets like those found in acoelans, leaving the status of ambulacrarian nervous systems unsettled.
A combination of complex brains and intelligence in the above-defined sense of higher cognitive abilities are found among ecdysozoan invertebrates in some orders of insects (e.g. in blattoids, dipterans, hymenopterans) and among lophotrochozoans in octopodid molluscs (cf. ), among vertebrates in some teleost taxa (e.g. cichlids), in corvid and psittacid birds, and among mammals in cetaceans, elephants and primates . Accepting the idea that the last common ancestor of protostomes and deuterostomes already possessed a tripartite brain of modest complexity, these brains have undergone a strong further increase in brain complexity and intelligent behaviour independently, because each of the major animal taxa mentioned had ancestors that revealed neither complex brains nor intelligent behaviour.
5. Insect brains and intelligent behaviour
As promostome taxa in general, most ecdysozoan taxa possess relatively simple nervous systems consisting of a nerve ring surrounding the oesophagus and often exhibiting lobes or ganglia, and a varying number of nerve cords, often with local ganglia. Many endoparasitic taxa show signs of strong secondary simplification . More complex brains are found in pan-arthropods (onychophorans, chelicerates, crustaceans, insects) as the result of the fusion of the first three ganglia .
The insect nervous system consists of a brain and ventral nerve cords with ganglia (cf. ). The brain is composed of a large protocerebrum, a smaller deutocerebrum and a very small tritocerebrum. Owing to the loss of antennae, chelicerates have no deutocerebrum. The protocerebrum consists of two hemispheres, which are continuous with the lateral optic lobes processing the input from the compound eyes. In the median protocerebrum, the corpora pedunculata or ‘mushroom bodies' (MB) are found .
MBs have been demonstrated to be the substrate of developed cognitive and social functions found in insects as well as in other arthropods. Among insects, complex MBs have evolved apparently independently in taxa of butterflies (Lepidoptera), dragonflies (Odonata, Anisoptera), cockroaches (Blattoidea) and hymenopterans (Hymenoptera, i.e. wasps, bees, ants, etc.). In cockroaches, flies, bees and wasps, among others, the MBs are very large and composed of two calyces, and a peduncle consisting of two lobes, the α and the β lobe. In the honeybee, the MBs occupy about half of the volume of the brain [9,10]. The somata of neurons, the ‘Kenyon cells' (KCs)—which in the honeybee number about 300 000 (R. Menzel 2010, personal communication) and whose axons (‘Kenyon fibres’) form the peduncle—are the smallest ones found among insects, and their packing density is 15 times higher than the highest ones found in the vertebrate brain. In the bee, the KCs receive visual, olfactory and somatosensory input from about 800 projection neurons via about 1 million presynaptic contacts, plus about 10 postsynaptic contacts. Axons of KC divide into the peduncle, and one collateral enters the α, and another one the β lobe. The calyces exhibit three regions, the lip ring region processing olfactory input, the collar ring region processing visual input and the basal ring region processing mixed olfactory and mechanosensory input. In hymenopterans as well as in some other insect taxa, the MB represents a highly complex multimodal centre that forms the neural basis of processing and integrating olfactory, somatosensory and visual information and enables complex cognitive functions and complex behaviour.
A large variety of insects including honeybees possessing complex mushrooms exhibit an impressive repertoire of behaviour in the domain of feeding, spatial orientation (‘navigation’), social and communicative behaviour, and can learn very quickly, especially the association between the colour and odour of flowers (for an overview, see [6,11,12]). This indicates a high degree of behavioural flexibility. This includes configural learning, contextual learning, i.e. to exert different kinds of behaviour depending on the site and the conditions, and categorical learning. Bees are able to learn the category ‘same–different’ and to transfer this concept to novel stimulus arrangements (cf. ). They master so-called delayed match-to-sample or its opposite, the delayed-non-match-to-sample.
The existence of such cognitive maps for spatial orientation has been debated for years (cf. ). Bees appear to possess ‘map-like’ information about the ground structure and use this map for return. In general, they appear to use spatial memories ‘opportunistically’ in three different contexts, i.e. (i) a general landscape memory acquired via initial orienting flights, (ii) route memory while flying repeatedly from and to a specific field location, and (iii) the dance memory while following dances .
The evolution of large and complex mushrooms has been interpreted as a consequence of the demands of increased sociality in the sense of the well-known ‘social brain hypothesis' developed in the context of primate–human brain evolution . However, a recent comparative study by Farris & Schulmeister  demonstrated that complex MBs are likewise found among non-social insects, including solitary and parasitoid hymenopterans as well as cockroaches, dragonflies and butterflies, all of which exhibit highly developed abilities of spatial learning (cf. , for cockroaches). The authors conclude that need for improved navigation and spatial learning rather than sociality was the major driving force for the evolution of complicated MBs.
6. Brain and intelligence in molluscs
As in ecdysozoans, in lophotrochozoans most taxa possess simple to moderately complex brains . Within endoparasitic platyhelminths (cestodes, nematodes), there is massive secondary simplification of the nervous system. In contrast, some predatory annelid polychaetes have multilobed cerebral ganglia with a protocerebrum containing mushroom-like structures. Among molluscs, gastropods have relatively simple cerebral ganglia, and bivalves show clear signs of secondary simplification. In contrast, cephalopods possess complex to very complex brains, which may have evolved several times independently . The most complex brains are found in squids (Theutidae) and octopods (Octopoda) .
The nervous system and brain of Octopus is the largest and most complex one among invertebrates [19,20]. It contains about 550 million neurons, 350 of which are located inside the eight arms, about 160 million neurons in the giant optic lobes and 42 million neurons inside the brain. The brain is divided into 16 lobes . It has a ventral motor portion consisting of several lobes that are involved in the control of feeding, locomotion and colour change, and a dorsal portion exerting sensory information processing and higher cognitive functions [21–23]. The brain receives visual afferents from the eyes and the subsequent optic lobes as well as tactile–chemosensory information from touch and taste receptors of the arms.
The vertical lobe is considered the most complex part of the Octopus brain [19,22,23]. It contains about 26 million neurons—more than half of the neurons inside the brain. It consists only of two major types of neurons, i.e. little more than 26 million tiny interneurons, the smallest inside the Octopus brain, and 65 000 large projection neurons, and the former converge on the latter. The vertical lobe receives visual afferents predominantly from the median superior frontal lobe. These afferents form a distinct tract composed of 1.8 million fibres, which terminate in the rind of the vertical lobe . Processes of the 26 million interneurons located there penetrate that tract in a rectangular fashion and form ‘en passant’ contacts with them. As major centre for ‘higher’ cognitive abilities of Octopus, the vertical lobe is closely connected, via the projection neurons, to the subvertical lobe, which contains about 800 000 neurons. The interaction of both lobes is based on the work of an impressively regular network of millions of crossing fibres.
The large and complicated brains found in squids and octopods correlate well with their cognitive abilities—presumably in the context of their predatory lifestyle [22,24]. Octopus not only remembers well where tasty food can be found, but after travelling far from home, it often returns home via the shortest route which it had never taken before. Such behaviour demonstrates that Octopus has a good spatial memory, but it is unclear whether it possesses a ‘mental map’, because some experts argue that the animal simply applies path integration. Other observations and experiments demonstrate that Octopus uses its siphon for cleaning its cave and its surroundings from sand and garbage. Also, it could be observed that the animal collects little stones and piles them in front of the entrance to their cave for protection against predators. Some Octopus specialists interpret this as evidence for tool use. Finally, these animals became famous for playing with plastic bottles and being able to unscrew lids from jars filled with prawn.
It has been reported that Octopus is capable of learning by pure observation of the behaviour of its conspecifics . While other experts were unable to reproduce these results, Fiorito & Chichery  published that removal of the vertical lobe abolished the ability for ‘learning by observation’ in Octopus. Until today, it is unclear how the findings by Fiorito & Scotto  should be interpreted, i.e. whether or not they give clear evidence for observational learning.
7. Deuterostomes, chordates and vertebrates
Among deuterostomes, echinoderms have two decentralized nervous systems: an ectoneural system with sensory functions and a hyponeural system with motor functions. Hemichordates posses a ventral and dorsal nerve cord connected by nerve rings in the head lobe and around the gut. It is debated whether these types of nervous systems found in ambulacrarians represent an ancestral or secondarily simplified state. The simple nervous system of urochordates may likewise be the result of secondary simplification. The cerebral vesicle of the cephalochordate Branchiostoma, containing 20 000 neurons, possesses all developmental genes required for the formation of the craniate–vertebrate brain .
Myxinoids and vertebrates generally have pentapartite brains consisting of a medulla oblongata, metencephalon, mesencephalon, diencephalon and telencephalon. The brains of galeomorph sharks, myliobatid rays, some teleost taxa (e.g. cichlids), corvid and psittacid birds and a number of mammals, particularly elephants, cetaceans and primates, represent the most complex ones.
(a) Hagfish and lamprey
Among chordates, hagfish (Myxinoidea) and lampreys (Petromyzontida) have the smallest relative brain size . Nothing is known about higher cognitive abilities in these groups.
(b) Cartilaginous fishes
Among cartilaginous fishes, i.e. sharks, rays and chimaeras, galeomorph sharks and myliobatid rays independently evolved large and complex brains, especially with respect to the size and anatomical organization of the telencephalon reaching relative proportions found in mammals. The so-called central nucleus appears to be the most important sensory convergence centre . Unfortunately, no systematic studies on cognition and intelligence of sharks and rays are available to date.
The structural and functional organization of the telencephalon of the actinopterygian bony fishes and possible homologies of its pallial regions with those of the other vertebrates is a matter of debate. At first glance, the dorsal telencephalon of the actinopterygians does not reveal any similarity with that of other vertebrates. The striking differences are explained by some authors by the fact that the telencephalon of the actinopterygians is an everted pallium, whereas the telencephala of the other vertebrate groups have an evaginated pallium. (cf. ). In the evagination type, the walls of the embryonic telencephalon thicken and bulge outwards encircling the lateral ventricle and divide into the mentioned pallial and subpallial parts. In the eversion type, the subpallial parts remain in a medial position along the midline of the telencephalon, whereas the pallial parts bend outwards and then downwards. As a consequence, the medial pallium of this everted telencephalon now occupies a lateral and increasingly ventral position. Accordingly, there is a new ‘medial zone Dm’, continued laterally by a ‘central zone Dc’ and a ‘lateral zone Dl’ and finally in a caudal position a ‘dorsoposterior zone Dp’. According to Wullmann & Vernier , this latter ’dorsoposterior zone’ receives olfactory input and therefore corresponds to the lateral olfactory pallium of the other vertebrates.
Thus, all these zones appear to represent standard parts of the pallium. Unclear is the homology of the ‘central zone Dc’ of the pallium. It gives rise to numerous pathways to the olfactory bulb and descending to various diencephalic and mesencephalic nuclear regions and, in some teleosts, via both the medial and lateral forebrain bundle, to the torus semicircularis and the cerebellum and valvula. Thus, the central zone is the main efferent station of the actinopterygian pallium.
Experts state that some teleost taxa, above all cichlids, exhibit cognitive abilities, which—in the eyes of the authors—are comparable to those of primates . Cichlids represent one of the behaviourally and ecologically most diverse groups of vertebrates. They are well known for having evolved rapidly within large lakes, especially the Great African Lakes, and many species evolved only 100 000 years ago .
According to Bshary et al. , in many cichlid species, there is individual recognition not only on visual, but also on purely auditory cues in the context of brood care behaviour. In some cichlid species with ranking orders, subordinate individuals often exhibit submissive or appeasement behaviour towards high-ranking group members in order to reduce aggression, and such appeasement behaviour is also shown during courtship. As in mammals and particularly in primates, cheating is answered by exclusion and punishment of the cheater, which means that the individuals are capable of remembering their partners' behaviour during past interactions. Also, teleosts (e.g. guppies) collect socially important information by observing the interactions between conspecifics. There are many reports about social learning by observation in the context of site preferences, food sources, anti-predator behaviour (e.g. mobbing an Octopus) and learning of young fish what to eat and what to avoid by observing adults. There are likewise many examples for cooperative hunting. Thus, it seems that in cichlids, the need for ‘social’ rather than for spatial or instrumental intelligence favoured the evolution of complex brains.
Amphibians generally have small brains in absolute as well as relative terms. Their relative brain weight overlaps with that of agnathans, bony fishes and reptiles [3,28]. Brain anatomy exhibits the signs of secondary simplification, most probably as a consequence of increased genome and cell size as well as consequent decrease of metabolic and developmental rates [33,34]. These phenomena are strongest in salamanders (Caudata) and caecilians (Gymnophiona) and weakest in frogs (Anura), as revealed by reduction or even absence of cell migration, nucleization and lamination (e.g. in the optic tectum and telencephalic pallium) .
While traditionally amphibians were regarded as being highly instinct-bounded or even mere ‘reflex machines' , extended studies in our laboratory demonstrated associative learning based on reward, punishment and omission of reward in frogs . In salamanders, behavioural and electrophysiological studies revealed a control system for visual attention resembling that of mammals [38,39].
‘Reptiles', i.e. a paraphyletic group consisting of turtles, tuataras, squamates and crocodiles, likewise have absolutely and relatively small brains, whereas the anatomy of their brains is distinctly more complex than that of amphibians in the sense of formation of distinct nuclei and degree of lamination, e.g. in the optic tectum and pallium . Like in amphibians, ‘higher’ intelligence has not yet been convincingly demonstrated in ‘reptiles'. Leal & Powell  reported that the tropical lizard Anolis evermanni exhibits behavioural flexibility across multiple cognitive tasks, including solving a novel motor task using multiple strategies and reversal learning, as well as rapid associative learning. The authors stated that this degree of behavioural flexibility found in Anolis is comparable with that commonly associated with bird or mammal species (see below). However, Vasconcelos et al. , in their critical review, argue that in the study of Leal and Powell , rather than reversal learning or other kinds of behavioural flexibility, the lizards exhibit simple instrumental conditioning.
In birds and mammals, the telencephalon including the pallium turns out to be the neurobiological substrate of high intelligence . However, at first glance, there are striking differences in the anatomy of the dorsal telencephalon of mammals and birds in the sense that in mammals there is a clear-cut distinction between a six-layered cortical region and a compact region of the striatopallidum as the telencephalic part of the basal ganglia, whereas the avian dorsal telencephalon gives the appearance of a compact cell mass without obvious lamination or other substructures. In previous times, this made neurobiologists believe that most of the avian dorsal telencephalon is a hypertrophied striatopallidum. On the basis of the pioneering work by Karten , the mentioned differences are explained today in the way that the avian dorsal telencephalon does not, indeed, form a six-layered cortex, but is composed from dorsal to ventral of compact layers consisting of a superficial hyperpallium (H), an intercalated hyperpallium (IH), a dorsal and ventral mesopallium (MD, MV), a nidopallium (N) including a central entopallium (E) and a ‘posterior field L’ as termination fields of visual and auditory afferents, respectively, an intercalated nidopallium (IN) and an arcopallium (AP) found at the caudoventral part of the nidopallium. Both nidopallium and arcopallium surround the striatopallidum.
According to the ‘nuclear-to-cortex theory’ of the bird pallium, IN and IH correspond to layer 4 neurons of the primary sensory cortex of mammals (see below), N and H to layers 2 and 3 of the secondary mammalian cortex, MV and MD to layers 3 and 6 of the tertiary cortex and AP to layer 5 of the motor cortex and of other cortices [42,44–46].
A special region of the pallium is the ‘nidopallium caudolaterale’. According to Güntürkün, functionally it strongly resembles the mammalian dorsolateral prefrontal cortex (dlPFC—see below), because it is involved in working memory, action planning, behavioural flexibility and creativity—in essence in ‘intelligence’ . Like the mammalian dlPFC, it is a multimodal convergence centre and receives a strong dopaminergic input from the ventral tegmental area and nucleus accumbens.
During the past two decades, it has become evident that some bird taxa, above all corvids and psittacids, possess cognitive abilities many of which appear comparable to those of primates. In nature as well as in captivity, corvids are shown to use natural objects as tools or to modify them until they have the right length or pass through an opening [48,49]. New Caledonian crows (Corvus moneduloides) spontaneously make tools out of screw pine (Pandanus) leaves. In that context, they fabricate strips of different size and shape for different purposes . According to Hunt et al. , crows occasionally make hooks in the wild, too, and the famous New Caledonian crow Betty had seen a hooked wire before, but this type of intentional tool-making is regarded as unique. Crows use short sticks to retrieve longer ones from a box, which was then used to retrieve food from a box (metatool)—a behaviour that otherwise has been observed in primates only .
In that context, the question arises whether corvid (or other) birds exhibit a causal understanding of tool use. Recent experiments by Taylor et al.  cast some doubts on such an assumption. The authors conducted standard string-pulling experiments with New Caledonian crows comparing experienced and naive animals. The animals had to pull up strings with meat at one end. When the animals had full sight of the string and the meat while pulling, naive ones solved the problem spontaneously in most cases. But when visual control of string-pulling was restricted, naive animals could no longer solve the problem spontaneously and only after extended trial and error learning, and the performance drastically dropped even in the experienced ones. According to the authors, this demonstrates that in the string-pulling experiment problem-solving is based primarily on reinforcement learning. Thus, the question of truly insightful problem-solving in corvids remains open.
Mirror self-recognition is considered a highly cognitive ability that has been found in apes, elephants and dolphins (see below). Recently, a team led by Güntürkün and co-workers  demonstrated that at least the common magpie (Pica pica), a corvid, passes the mark test. When the plumage of the magpies was marked below the beak, i.e. at a site that could not be inspected without the mirror, the animals started cleaning themselves and trying to touch the spot. They did not respond to pictures, padded or alive, marked or unmarked magpies behind a glass pane, i.e. they did not confound their own mirror image with that of conspecifics. Magpies are likewise highly social animals and exhibit an unusual ability to recognize and relocate cached objects (food, but also glittering objects). They are capable of recognizing conspecifics and other animals individually. This adds another case to the unusual cognitive abilities of corvid birds.
The ability for mind-reading was reported for birds in a food caching experiment. A raven was able to take into account what an other raven has seen or not seen, while a human experimenter was hiding food in a cache. Ravens were quicker at pilfering the human-made caches when facing a fully informed raven competitor compared with a partially or non-informed one . The author interprets his finding as a ‘precursor step to a human-like understanding of the others' mind'.
A widely known example of avian high intelligence is Alex, an African gray parrot trained and studied by Pepperberg . According to Pepperberg, Alex could identify 50 different objects and recognize quantities up to six, he could distinguish seven colours and five shapes and understand the concepts of ‘bigger’, ‘smaller’, ‘same’ and ‘different’, and that he was learning ‘over’ and ‘under’. He had a vocabulary beyond 100 words and gave the impression of understanding what he said. He could form concepts of form, colour, objects and to a limited degree had an understanding of syntax in language. He had the ability to count objects up to six even verbally, chose certain objects on command and had an understanding of ‘zero’.
High intelligence that characterizes monkeys and apes, including humans, is closely linked to the size and neuronal complexity of the isocortex [28,55], which—with the exception of insectivores and cetaceans—is six-layered (as opposed to most of the limbic cortex, which is three- to five-layered) . Among cortical areas, the dorso- and ventrolateral prefrontal cortex (PFC) appear to be the neurobiological basis of so-called general intelligence in mammals [56,57]. It is involved in the comprehension and processing of the temporal–spatial structure of sensory information and cognitive mental events such as thinking and imagining, predominantly in the context of action planning and action preparation, but also problem-solving and decision-making. It is also an important component of working memory.
The dorsolateral PFC of primates receives input mainly from the posterior parietal cortex regarding position and movement of head, neck, face and hands, as well as information about spatial orientation and spatial aspects of action planning. The ventrolateral PFC, in contrast, receives input mainly from the temporal lobe carrying complex visual and auditory information, e.g. in the context of the meaning and relevance of objects and scenes, as well as language-related information from the left temporal lobe (mostly from the Wernicke language centre). In the ventrolateral PFC, we find the Broca language centre, which is involved in grammatical and syntactical aspects of language .
In the following, I will discuss a number of cognitive abilities commonly cited as evidence for high primate intelligence, i.e. tool use, quantity representation, extended working memory, imitation, mirror self-recognition, theory of mind and conscious attention (cf. ).
(i) Tool use
Tool use is commonly found among primates. Ring-tailed lemurs (Lemur catta) have been reported to successfully manipulate a puzzle feeder in the wild , which is the only known case of lemur tool use in the wild. In captivity, however, manipulatory skills of lemurs with novel objects are roughly comparable with those of some New and Old world monkeys. The gray mouse lemur (Microcebus) mastered opening boxes in different ways, and aye-ayes (Daubentonia) demonstrated basic understanding of features of tools by solving a can-pulling task . Systematic tool use including limited forms of tool-making is found in the capuchin monkey (Cebus) [60,61].
Chimpanzees are known to fabricate and use a wide range of complex tools, and have been shown to vary in their tool use at many levels, for example preparing twigs for ant and termite dipping . Tool kits consist of about 20 types of tools for various functions. Only chimpanzees appear to be able to use one type of raw material to make different kinds of tools, or to make one kind of tool from different raw materials. They use tool sets in a sequential order, make use of composite tools and combine tools to a single working unit [63,64]. In this context, chimpanzees as well as orangutans exhibit insightful problem-solving [65,66]. They are engaged in action planning, mentally pre-experience an upcoming event and are able to select objects needed for much-delayed future tool use.
(ii) Quantity representation
Lemurs are capable of controlling their impulsive gesture towards a larger option, when selection of a smaller quantity of food is rewarded with a larger one. They also learned to associate a graphic representation of the reward with the corresponding quantity, even though only one subject consistently selected the representation of the smaller quantity to be rewarded with the larger quantity of food . The fundamentals of abstraction appear to be present in prosimians. Nevertheless, numerical discrimination is superior in monkeys and apes. Capuchin monkeys are able to judge larger quantities of two sets contrasting up to five items in food-choice experiments . Quantity-based judgements for two sets with up to 10 items were tested in rhesus monkeys (Macaca mulatta), gorillas, chimpanzees and bonobos. Rhesus monkeys selected the larger of the two sequentially presented sets reliably when one set had fewer or more than four items, whereas great apes did so, even when the quantities were large and the numerical distance between them was small .
(iii) Working memory
One much-used task for testing working-memory abilities is the delayed-matching-to-sample task (DMTS) or the complementary form of delayed-non-matching-to-sample. In DMTS, the subject is shown a rewarded stimulus and has to keep it in mind for a variable period of time during which the stimulus is not visible anymore, and then identify it out of a pair of stimuli. Macaques master a delay of 2–9 min after long training. Dolphins reach a maximum of 4 min. In humans, the capacity of working memory is substantially increased by the ‘phonological loop’ [70,71], but when human subjects are prevented from talking to themselves by inner speech or loudly, then they are no better than dolphins and macaques.
Some years ago, Inoue & Matsuzawa  reported the astonishing capacity for numerical recollection in chimpanzees (three mother–offspring pairs). In a numerical sequence task, the animals had to learn the sequence of Arabic numerals from 1 to 9, which appeared in different on-screen positions. Then the numerals were replaced by white squares. The subjects had to remember which numeral appeared in which location, and then touch the white squares in the correct sequence. All naive animals mastered this task, but the performance of the three young chimpanzees was always better than that of the three mothers. Humans were slower than the three young chimpanzees. It would be interesting to test young human children under the same conditions.
Imitation was long considered an inferior kind of learning in the sense of meaningless copying of a certain behaviour. Only in recent years has it become clear that imitation is a higher-order cognitive ability. However, to date, there is no universally accepted definition of imitation, and some kinds of behaviour previously seen as imitation are now interpreted differently. One of these kinds of imitation-like behaviour is response facilitation or emulation found in a wide range of animals, which means that seeing an action ‘primes' the individual to do the same, and the individual, by trial and error, finds the same or a very similar solution of the problem .
Of importance is the social component in emulation. For example, young baboons (Papio) quickly learn which kinds of fruit are edible after one group member has tasted a fruit. Vervet monkeys (Chlorocebus), likewise Old World monkeys, learn this task more slowly, although they live in the same environment as baboons. The explanation for the difference may be that young baboons have a close social life and show great interest in each other, whereas this is not the case for the Vervet monkeys.
Imitation of human behaviour is frequently found among apes, e.g. in orangutans (Pongo pygmaeus). In the Indonesian Tanhung Putting National Park, orangutans were observed imitating everyday actions of the human park personnel . In some cases, the animals seemed to understand the sense of the copied action, in other cases, they carried out the actions ‘just for fun’. This included the decanting of fuel from a barrel into a canister, sweeping trails, making fire, using a saw, mixing ingredients for a pancake or dish-washing.
Chimpanzees and other great apes show imitative abilities beyond those of other primates. The recent view is that great apes are capable of mentalizing about others and have some understanding of intentionality and causality. Chimpanzees are able to distinguish between an experimenter who is either unwilling or unable to give them food. Hence, they do not simply perceive the behaviour of others, but also interpret it . Capuchin monkeys were shown to distinguish between intentional agents and unintentional objects .
(v) Mirror self-recognition
Gordon Gallup was the first to demonstrate that—besides humans—at least some chimpanzees and orangutans are capable of recognizing themselves in the mirror, but only in less than half of animals tested and not always in those that pass the test. It is mostly the young animals that display mirror self-recognition, and even they rapidly lose interest in such experiments .
(vi) Theory of mind: understanding the others
The question whether or not non-human animals possess a theory of mind (ToM), i.e. the ability to understand another individual's mental–emotional state, is hotly debated, as is the related ability to ascribe to a conspecific a certain knowledge or false knowledge (or false belief) and to take both into account in the planning of own behaviour: in the 1980s and 1990s, Povinelli et al. [77,78] conducted experiments showing that at least some chimpanzees possessed the ability to attribute certain knowledge to other chimpanzees or humans and to take into account that knowledge in their own behaviour. Later, however, Povinelli and Vonk became sceptical and could find no convincing evidence for the existence of a ToM and knowledge attribution in chimpanzees or other animals, but argued that their own findings could be better explained as the result of operant conditioning . However, other primatologists strongly disagree and point to substantial drawbacks in the method applied by Povinelli and co-workers. Tomasello et al.  believe that chimpanzees possess at least some aspects of a ToM and knowledge attribution comparable to that of human children at an age of 3 to 4 years.
(vii) Conscious attention
After Wolfgang Köhler's pioneering work with chimpanzees on Tenerife during the World War I, my academic teacher Bernhard Rensch from the University of Münster, Germany, was among the very first who carried out experiments to assess the cognitive–mental abilities of chimpanzees, and his favourite subject was Julia. In a typical experiment, Julia had to draw an iron ring out of a wooden maze covered with a glass plate using a magnet. She could choose between two alternative paths, of which only one led out of the maze, and she had only one move. Rensch & Döhl  started with simple mazes, but eventually confronted Julia with rather complex ones, which we humans can master only after carefully ‘wandering’ with our gaze through the maze. Julia did exactly that and in most cases (86%) chose the right path. Rensch  interpreted these finding as clear evidence that at least chimpanzees possess conscious awareness and can solve problems mentally.
8. Brains and intelligence in elephants and cetaceans
Among mammals, Indian and African elephants, whales and most dolphins have brains that are larger to much larger than the human brain. Elephant brains may have weights of up to 6 kg, and the false killer whale (Pseudorca crassidens) up to 10 kg. Therefore, one would expect that their intelligence is superior to that of humans. However, controlled experiments on the alleged super-intelligence of the dolphins yielded somewhat mixed and often disappointing results.
Elephants possess a magnificent spatial orientation that enables them to head for water ponds as distant as 60 km. Likewise, they can recognize human individuals after decades, as Rensch & Altevogt  have demonstrated. All this stands in sharp contrast to their rather unimpressive cognitive abilities [84,85]. As to learning abilities, Rensch & Altevogt  report the cumbersome procedure of carrying out a simple operant conditioning experiment teaching an elephant to distinguish between black and white or small and large objects. After a number of failures, Plotnik and co-workers  demonstrated mirror self-recognition in at least one out of three Indian elephants, Elephas maximus.
Although dolphins are able to distinguish between objects differing in shape, they are incapable of categorization, for example of distinguishing between ‘round’ and ‘triangular’ objects and assigning an unfamiliar round or triangular object to one of these two categories—something that pigeons, crows, parrots, dogs, any kind of primates and even bees are capable of .
Reiss & Marino  succeeded in demonstrating that captive-born bottleneck dolphins (Tursiops truncatus) are capable of mirror self-recognition. At the beginning, the dolphins showed great interest in the marks attached to their bodies, which they could not inspect without the help of the mirror, but like the chimpanzees and unlike young (as well as older) humans, they rapidly lost interest in the procedure. Possible neurobiological reasons for the mentioned differences in cognitive abilities of elephants and dolphins compared with primates are discussed in the twin article by Dicke & Roth .
I have tried to demonstrate that high intelligence, as defined at the beginning, has evolved at least several times independently in some insect taxa, particularly in hymenopterans, and in octopodids molluscs, as well as in some teleost taxa (e.g. cichlids). The last common ancestor of both ecdysozoans and lophotrochozoans, as well as the last common ancestor of each group, may either have had a non-centralized nerve net (or two of them, an epidermal and a hypodermal one) or a tripartite, though simple brain, which then in very many taxa underwent additional simplification. In any case, the highly complex brains found in insects and in cephalopods must have evolved independently, and with them high intelligence.
Among craniates–vertebrates, the lowest levels of brain complexity and intelligence are found in hagfish and lamprey, followed by amphibians. Likewise, ‘reptiles' in the traditional sense (i.e. turtles, snakes and lizards) are not famous for superior intelligence, and their brains are definitely less complex than those of birds and mammals. However, among teleosts, there are taxa like cichlids that are believed to exhibit signs of ‘primate-like’ intelligence.
With respect to intelligence, among birds, parrots and corvids stand out , which interestingly are all vocal learners , and among mammals primates stand out [3,55]. Within primates, prosimians exhibit manipulatory, perceptual and cognitive capacities, although often only as a basic ability. New World monkeys possess moderate capacities in various cognitive domains that partially overlap with those of Old World monkeys. The behaviour of the latter shares characteristics with apes, although great apes clearly outperform the other non-human primate taxa in most respects, with humans on top. Despite their very large brains, elephants and cetaceans turn out to possess an intelligence definitely inferior to monkeys and apes and also possibly to corvid and psittacid birds.
There is much speculation about differences in the ‘driving forces' with respect to increases in brain size and complexity as well as cognitive abilities (cf. ). On the basis of the data presented in this review, needs for spatial learning and foraging strategies in insects and cephalopods, for social learning in cichlids, for instrumental learning and spatial orientation in birds and social as well as instrumental learning in primates appear to be the major forces.
I have no competing interests.
I received no funding for this study.
I thank two anonymous reviewers for helpful comments and Prof. Ursula Dicke, University of Bremen, for critical reading of the manuscript.
One contribution of 16 to a discussion meeting issue ‘Origin and evolution of the nervous system’.
- Accepted July 31, 2015.
- © 2015 The Author(s)