In this article, we summarize the traditional view of telencephalic evolution before reviewing more recent findings and insights. We then present the new nomenclature that has been developed by the Avian Brain Nomenclature Forum, and discuss its implications for our understanding of vertebrate brain evolution and its associated homologies.

The classical view of telencephalic evolution, which is still prevalent in classrooms and textbooks, began in the late nineteenth and early twentieth centuries after the publication of The Origin of Species by Darwin². Inspired by Darwin’s theory, between 1885 and 1908 Edinger formulated an influential, evolution-based model of brain organization^1,3,4. Edinger and other early comparative neurobiologists combined Darwin’s concept of ‘evolution’ with the nineteenth-century version of Aristotle’s ‘scala naturae’, which resulted in the view that evolution was progressive and unilinear⁵ — from fish, to amphibians, to reptiles, to birds and mammals, to primates and, finally, to humans — ascending from ‘lower’ to ‘higher’ intelligence in a chronological series. They believed that the brains of extant vertebrates retained ancestral structures, and, therefore, that the origin of specific human brain subdivisions could be traced back in time by examining the brains of extant non-human vertebrates. In making such comparisons, they noted that the main divisions of the human CNS —the spinal cord, hindbrain, midbrain, thalamus, cerebellum and cerebrum or telencephalon — were present in all vertebrates (FIG. 1a). Edinger, however, noted that the internal organization of the telencephala showed the most pronounced differences between species. In mammals, the outer part of the telencephalon was found to have prominently layered grey matter (FIG. 1b, green) whereas the inner part had nuclear grey matter (FIG. 1b, purple). The inner part was located ventrally to the lateral ventricle. The outer part was more elaborate and folded in humans than in smaller mammals. In non-mammals, the outer and inner parts of the telencephala were mainly composed of nuclear grey matter, most of which was located ventrally to the lateral ventricle in reptiles and birds (FIG. 1b, purple).

Figure 1 Avian and mammalian brain relationships

On the basis of these considerations, Edinger proposed that telencephalic evolution occurred in progressive stages of increasing complexity and size, culminating with the human cerebrum. He suggested that the stages proceeded in a ventral-to-dorsal direction, with each new vertebrate group acquiring a more advanced cerebral subdivision, much as the earth’s geological strata formed over time. He proposed that, first, there was the old brain, the palaeoencephalon (also called the basal ganglia or subpallium at the telencephalic base), which controlled instinctive behaviour, followed by the addition of a new brain, the neoencephalon (also called the pallium or mantle at the top of the telecephalon), which controlled learned and intelligent behaviour⁴. He, Ariëns Kappers and others named the telencephalic subdivisions within each vertebrate group with the prefixes ‘palaeo’ (oldest),‘archi’ (archaic) and ‘neo’ (new) to designate the presumed relative order of evolutionary appearance of each subdivision. In Greek,‘archi’ means the oldest, the first, or the most primitive, whereas ‘palaeo’ means ancient, primitive or old, but not necessarily the oldest. Both Edinger and Ariëns Kappers misinterpreted the meaning of these prefixes and reversed them, naming structures with ‘palaeo-’ to indicate the oldest or first and ‘archi-’ to indicate old. They added to these prefixes the root word ‘striatum’ for the presumed palaeoencephalic subdivisions and ‘pallium’ or ‘cortex’ for the presumed neoencephalic subdivisions^1,4,6–8. The term ‘striatum’ was used because a large part of the basal ganglia (palaeoencephalon) in mammals, now commonly called the caudate–putamen, has fibre bundles coursing through it that give it a striated appearance.

The classical view that became dominant was that the primordial telencephalon of fishes had a relatively small pallium and a larger subpallium, both of which were entirely devoted to olfactory information processing. The fish subpallium was named ‘palaeostriatum’ (old striatum), and was thought to be the antecedent of the human globus pallidus (FIG. 1b). Amphibians were thought to have evolved an ‘archistriatum’ (archaic striatum) above the palaeostriatum, which was proposed to be the antecedent of the human amygdala. Reptiles were thought to have evolved a ‘neostriatum’ (new striatum) above the archistriatum, which was proposed to be the antecedent of the human caudate and putamen. The palaeostriatum of reptiles was also thought to have elaborated into an older part (primitivum) and a newer part (augmentatum), both of which were considered homologous to the human globus pallidus. Following this, birds were thought to have evolved a large additional basal ganglia subdivision, the ‘hyperstriatum’ (hypertrophied striatum), which was considered to be unique to birds⁹.

The fish pallium was named ‘palaeocortex’, and was proposed to be the antecedent of the human olfactory cortex. Reptiles were thought to have evolved an ‘archicortex’, also thought to be olfactory and primitive, that was said to be the antecedent of the human hippocampus. Birds were thought not to have evolved any further pallial regions. By contrast, mammals were thought to have evolved the latest and greatest achievement, a ‘neocortex’, from the palaeocortex and/or archicortex⁶. The archicortex and/or palaeocortex, with their 2–3 cell layers, were assumed to be primitive; the neocortex, with its 6 layers, was assumed to be more recently evolved and a substrate for more sophisticated behaviour.

There were dissenting voices to the classical view^10–12. Some of its proponents also made partial or tentative retractions^13,14. However, alternative views were not widely embraced. Instead, the classical view was codified in the important 1936 comparative neuroanatomy text by Ariëns Kappers, Huber and Crosby¹⁴ and became pervasive throughout neuroscience.

Substantive challenges to the classical view of the subpallium began in the 1960s and 1970s with the advent of new methods for determining both nervous system connectivity and the anatomical profiles of gene products⁵. These studies found that, in mammals, acetyl-cholinesterase enzymatic activity was enriched in the neostriatum¹⁵. In birds, high acetyl-cholinesterase activity was found only in the palaeostriatum augmentatum and associated lobus parolfactorius (LPO)^15,16 (the LPO was considered to be part of the palaeostriatum augmentatum by Ariëns Kappers et al.¹⁴, but was named as a separate region by Karten and Hodos¹⁷). Other studies found that the mammalian neostriatum was highly enriched with dopaminergic terminals, which originated from midbrain neurons in the substantia nigra pars compacta¹⁸. In birds, again, only the palaeostriatum augmentatum and LPO were enriched with dopaminergic terminals, and this input originated from neurons in the midbrain^19,20. During the following decades, using new methods in double-label immunohistochemistry and tract tracing, the mammalian neostriatum was found to be enriched in two types of neuron: those containing the neuropeptide substance P (SP), which project to the internal part of the globus pallidus and substantia nigra, and those containing the neuropeptide enkephalin (ENK), which project to the external part of the globus pallidus^21–24. In birds, SP and ENK neurons are enriched in the palaeostriatum augmentatum (including the LPO)^24,25, and, like the equivalent neurons in mammals, project to different cell types within the adjacent avian palaeostriatum primitivum. In both birds and mammals, the SP neurons seem to be involved in promoting planned movement, whereas the ENK neurons seem to have a role in inhibiting unwanted movement. Further functional studies revealed that both the mammalian neostriatum and the avian palaeostriatum augmentatum (including the LPO) participate not only in instinctive behaviour and movement, but also in motor learning^26,27.

These apparent relationships between the subpallia of mammals and birds have been supported by molecular embryology studies^24,28–31. The developing subpallium in birds and mammals consists of two separate histogenetic zones that express different sets of transcription factors: a dorsal zone, which, in mammals, corresponds to the lateral ganglionic eminence and selectively expresses the transcription factors DLX1 and DLX2 but not NKX2.1; and a ventral zone, which, in mammals, corresponds to the medial ganglionic eminence and selectively expresses all three transcription factors. In mammals, the lateral ganglionic eminence gives rise to the dorsal striatum (neostriatum) and the ventral striatum (nucleus accumbens and part of olfactory tubercle). The homologous developing territory in birds gives rise to the structures that were previously called the palaeostriatum augmentatum (including the LPO) and the olfactory tubercle. The medial ganglionic eminence in mammals gives rise to various pallidal cell groups, including the dorsal pallidum (the globus pallidus) and the ventral pallidum. The homologous developing territory in birds gives rise to the structures that were called the palaeostriatum primitivum and ventral palaeostriatum. These avian and mammalian striatal and pallidal relationships are further supported by studies of the comparative expression patterns of more than 30 other genes in adult birds and mammals^32–36. Similar striatal and pallidal territories have been found in the so-called palaeostriatal regions of reptiles^37–40. Together, these studies indicate that the avian palaeostriatum augmentatum is homologous to the mammalian neostriatum and that the avian palaeostriatum primitivum is homologous to the mammalian globus pallidus.

With these challenges to the classical view of the subpallial relationships among birds, reptiles and mammals came challenges to the classical view of the relationships among their pallia. The mammalian pallium includes the areas known as palaeocortex, archicortex and neocortex; and has been said, more recently, to include both the claustrum and lateral parts of the amygdala^28,41,42 (FIG. 1c; Holmgren¹¹ originally proposed that the claustrum and part of the amygdala were pallial, but this view was largely ignored at the time). In birds, the finding that the structures that had been called hyperstriatum, neostriatum and archistriatum were neither striatum nor pallidum raised the question of which telencephalic sector these regions did represent. The results that were needed to answer this question also began to appear in the mid-1960s from pathway tracing^16,43–46 and behavioural studies^47–51. These studies found that the so-called avian neostriatum and hyperstriatum receive visual, auditory and somatosensory input from the thalamus, as does the mammalian neocortex. These avian brain regions also carry out the same type of sensory information processing as is performed by the mammalian neocortex. The so-called avian hyperstriatum accessorium and the archistriatum give rise to important descending projections to the premotor and motor neurons of the brainstem and spinal cord, like those of the mammalian cortico-bulbar and cortico-spinal pathways^43,52–54. Finally, like the mammalian neocortex, these avian brain regions carry out crucial roles in motor control and sensorimotor learning^55–67.

The apparent pallial relationships between these mammalian and avian brain regions were also supported by molecular embryology studies^28,68,69. During development, both the avian hyperstriatum and neostriatum and the mammalian pallium express the pallium-specific transcription factors EMX1, PAX6 and TBR1. The developmental data led to uncertainties about how much of the archistriatum is pallial^28,30. However, comparisons of the expression of the brain-derived neurotrophic factor (BDNF ) and the glutamate receptor mGluR2 in adult birds and mammals indicated that the entire avian archistriatum, as defined in brain atlases^17,70, expresses these pallium-specific mRNAs^34,36. Further studies of the comparative expression patterns of other glutamate receptors in adult birds and mammals³⁶ support these conclusions. Together, these studies indicate that the avian hyperstriatum, neostriatum, and archistriatum might be homologous to mammalian pallial regions.

This developing view was accompanied by several new proposals about one-to-one homologies between specific avian and mammalian pallial subdivisions. We will consider these in two groups — nuclear-to-layered hypotheses and nuclear-to-claustrum/amygdala nuclei hypotheses.

Both the above hypotheses have their limitations. For the nuclear-to-layered hypotheses, developmental studies have not been conducted to investigate whether the three types of serially connected neuron in birds arise from cell types similar to those that give rise to the cortical layers in mammals. Furthermore, not all gene expression patterns support one-to-one molecular relationships between avian pallial subdivisions and mammalian cortical layers⁷⁸. In addition, not all findings support the nuclear-to-claustrum/amygdala hypothesis. Although initial studies²⁸ reported that this hypothesis was supported by the lack of pallial EMX1 expression in the avian neostriatum (so-called ventral pallium) and the mammalian ventral claustrum and lateral anterior amygdala, recent fate mapping showed that the ventral claustrum contains dispersed EMX1-expressing cells and the lateral amygdala contains many EMX1-expressing cells⁷⁹. The antiquity of the claustrum has also been debated. One study⁸⁰ reported that monotremes (platypus and echidnas) lack a claustrum, whereas a later study⁸¹ reported that echidnas, but not platypus, have a rudimentary claustrum located more ventrally in the white matter relative to the location of the claustrum in other mammals. So, it is possible that some monotreme groups have lost the claustrum; that the echidna has independently evolved it or another deep cortical derivative; or that the claustrum is not an ancestral mammalian trait. Further investigation is required.

Despite an extensive revision of our understanding of telencephalic evolution, the common nomenclature used for the avian telencephalon has, until 2004 (REF. ⁸²), retained all of the classical evolution- and scala naturae-based terminology. For this reason, findings in ‘birdbrains’ have been habitually misinterpreted by neuroscientists studying non-avian brains as pertaining to the basal ganglia or as largely irrelevant to mammals. To rectify this problem, an international consortium of specialists in avian, mammalian, reptilian and fish neurobiology — the Avian Brain Nomenclature Consortium (BOX 1) — assembled with the goal of revising the terminology for the avian brain. Through online discussions, an Avian Brain Nomenclature Exchange web site, various meetings held over a period of 6 years and an Avian Brain Nomenclature Forum held at Duke University, North Carolina⁸³, the group developed a new terminology that represents the current understanding of avian telencephalic organization and its homologies with mammals⁸² (FIG. 1c). On the basis of the evidence summarized above, we concluded that the avian telencephalon is organized into three main, developmentally distinct domains that are homologous in fish, amphibians, reptiles, birds and mammals: pallial, striatal and pallidal domains (FIG. 1c). We renamed the subdivisions within each of these domains in birds with homology-based terms or roots that allow reference to named regions in mammals, and eliminated all phylogeny-based prefixes (palaeo-, archi- and neo-) that erroneously implied the relative age of each subdivision.

With this new understanding of the avian telencephalic organization and its homologies with that of mammals, we can generate more informed hypotheses and conclusions about telencephalic evolution in vertebrates. It is now apparent that the organization of the true basal ganglia among birds, mammals and other vertebrates (that is, distinct nuclear striatal and pallidal domains with more dopaminergic input into the striatal domain²⁴) is quite conserved. By contrast, the organization of the pallial domains of these groups is more varied. The avian hyperpallium has a unique organization that has so far been found only in birds⁶⁹. This consists of semi-layered subdivisions, and might have evolved more recently than the mammalian six-layered cortex, as birds evolved ~50–100 million years after mammals^85,86 (FIG. 2). The DVR (which, in birds, contains the mesopallium, nidopallium and arcopallium) is a nuclear, grey matter formation that is unique to birds and reptiles. The six-layered cortex is unique to mammals, and, as all the main groups of living mammals (monotremes, marsupials and placentals) have a six-layered cortex⁸⁷, it was presumably inherited from their common therapsid ancestor more than 200 million years ago (FIG. 2). Furthermore, new findings indicate that mammals did not arise from reptiles, but from therapsids, and that the last common ancestor of the reptile and mammal lineages was the stem amniotes⁸⁶. As all non-mammalian therapsids are now extinct (FIG. 2), it is difficult to trace from stem amniotes to mammals the evolutionary history of mammalian telencephalic organization — layered, nuclear or otherwise. Therefore, the reptilian nuclear pallial organization cannot be assumed to represent the ancestral condition for mammals.

Figure 2 Simplified modern view of vertebrate evolution

Further, it is now known that evolution is not invariably progressive or linear, so there is no basis for the view that more recently evolved species or structures are more advanced. In support of this conclusion, we now know that the telencephala of fishes are not devoted mainly to olfactory function, as the olfactory area represents only a limited portion of the fish pallium⁸⁸. In addition, fishes have a hippocampus (archicortex), and the main function of the hippocampus in fishes, reptiles, birds and mammals alike is not olfaction, but memory formation and spatial mapping^89,90.

So, as for birds, it might be best to abandon the use of the terms archicortex, palaeocortex, archistriatum, palaeostriatum and neostriatum for mammals and other vertebrates in favour of the alternatives — hippocampus or hippocampal cortex, piriform cortex, amygdala, striatum and pallidum. However, alternative terms for ‘neocortex’, such as ‘isocortex’, have not been universally accepted. ‘Neocortex’ would be appropriate if taken to refer to the uniqueness of this cortical structure among vertebrates. However,‘neocortex’ should not be taken to mean that it is the only unique form of pallial organization, that it evolved out of a palaeo- and/or archicortex, or that it is the newest pallial organization to have evolved. In the absence of a universally accepted alternative to neocortex, for the remainder of this article we use the term ‘six-layered cortex’. Although some regions of this cortical domain have fewer than six layers, we justify our use by analogy to the term tetrapods, which refers to all taxa derived from ancestral four-footed vertebrates, including snakes and whales.

Overall, the evidence indicates that there are pallial, striatal and pallidal domains in most or all vertebrates²⁴. Therefore, it is reasonable to propose that the telencephala of early fishes possessed all three domains, which were then inherited as a package by later vertebrates, and independently modified in them. The conserved organization of striatal and pallidal domains indicates that there might be constraints on how the basal ganglia can be organized. The diverse organizations of the pallial domains indicate that there are fewer constraints on how the pallium can be organized. This view has important implications for our understanding of neural mechanisms of cognition.

On the basis of this new understanding of avian brain organization and its evolutionary relationships, we estimate that, as in mammals, the adult avian pallium comprises about 75% of the telencephalic volume (FIG. 1c; calculated from sagittal series of pigeon and zebra finch brain sections). This realization of a relatively large and well developed avian pallium that processes information in a similar manner to mammalian sensory and motor cortices sets the stage for a re-evaluation of the cognitive abilities of birds, which, since the 1950s, have been increasingly appreciated as far more complex than was originally presumed^91,92. For example, pigeons can memorize up to 725 different visual patterns⁹³, learn to categorize objects as ‘human-made’ versus ‘natural’⁹⁴, discriminate cubistic and impressionistic styles of painting⁹⁵, communicate using visual symbols⁹⁶, rank patterns using transitive inferential logic⁹⁷ and occasionally ‘lie’^98,99. New Caledonian crows make tools out of leaves or novel human-made material, use them appropriately to retrieve food and are thought to pass this knowledge on to other crows through social learning^100,101. Magpies develop an understanding of object constancy at an earlier relative age in their lifespan than any other organism tested and can use this skill to the same extent as humans¹⁰². Scrub-jays show episodic memory — the ability to recall events that take place at a specific time or place, which was once thought to be unique to humans¹⁰³. This same species modifies its food-storing strategy according to the possibility of future stealing by other birds and, therefore, exhibits a behaviour that would qualify as theory-of-mind¹⁰⁴. Owls have a highly sophisticated capacity for sound localization, used for nocturnal hunting, that rivals that of humans and that is developed through learning⁶⁶. Parrots, hummingbirds and oscine songbirds possess the rare skill of vocal learning¹⁰⁵. This trait is a prerequisite in humans for spoken language and, with the exceptions of cetaceans and possibly bats, is not found in any other mammal¹⁰⁶. In addition, parrots can learn human words and use them to communicate reciprocally with humans. African grey parrots, in particular, can use human words in numerical and relational concepts^107,108, abilities that were once thought to be unique to humans.

So, many birds have cognitive proficiencies that are quite sophisticated, and some birds and mammals have cognitive proficiencies that clearly exceed all other birds or mammals. As these cognitive functions are carried out by the six-layered cortex in mammals¹⁰⁹ but by nuclear pallial areas in birds^{58,66,110,111}, it is clear that the mammalian six-layered cortical architecture is not the only neuroarchitectural solution for the generation of complex cognitive behaviours. It is also clear that pallial–cortical folding is not required. Birds apparently cannot use cortical folding because of the nuclear organization of their telencephalon; among mammals such folding is related more to absolute brain size than to behavioural complexity^112,113. The presence of specific brain subdivisions and connections are more important factors for the generation of behavioural complexity.

An example of how avian pallial and sub-pallial areas can interact to produce complex behaviour in the context of the new view of avian brain organization can be seen in the brain pathways that control learned vocal communication (FIG. 3). Most of the telencephalic auditory processing areas are in the pallium, adjacent to a smaller auditory area in the striatum (FIG. 3a). Likewise, most of the telencephalic vocal control nuclei are in the pallium, with one vocal nucleus in the striatum (FIG. 3b). The vocal nuclei that are involved in the production of learned vocalizations, including human speech in parrots¹¹¹, make up a pathway that directly innervates brainstem motor neurons (FIG. 3b, black arrows). This vocal motor pathway is similar to mammalian motor corticobulbar pathways¹⁰⁶. The vocal nuclei that are involved in the imitation of vocalizations form a pallial–basal ganglia–thalamic–pallial loop (FIG. 3b, white arrows). This vocal learning pathway is similar to mammalian cortical–basal ganglia–thalamic–cortical loops^27,106,114, which are involved in motor learning, sensorimotor integration and addictive behaviours. Other avian sensory and motor systems that are used for cognitive behaviours share a common circuit organization with the auditory and vocal pathways^63,64.

Figure 3 Auditory and vocal pathways of the songbird brain within the context of the new consensus view of avian brain organization

The inaccurate evolution-based terminology for the vertebrate brain that was used throughout the twentieth century became a severe impediment to the communication of scientific discoveries and the generation of new insights. Many of the tenets on which this old view of vertebrate telencephalic evolution was based have been refuted. The problems created by this view and its associated nomenclature have now been rectified for the avian brain with a new terminology that reflects the current understanding of vertebrate brain organization, homologies, evolution and function. This new understanding should facilitate a better assimilation of scientific insights into brain function through the study of birds.

This effort was supported by grants from the National Institutes of Health and National Science Foundation, and by the National Science Foundation Waterman Award to E.D.J.