Comparative Functional Anatomy of Marmoset Brains

Abstract

Marmosets and closely related tamarins have become popular models for understanding aspects of human brain organization and function because they are small, reproduce and mature rapidly, and have few cortical fissures so that more cortex is visible and accessible on the surface. They are well suited for studies of development and aging. Because marmosets are highly social primates with extensive vocal communication, marmoset studies can inform theories of the evolution of language in humans. Most importantly, marmosets share basic features of major sensory and motor systems with other primates, including those of macaque monkeys and humans with larger and more complex brains. The early stages of sensory processing, including subcortical nuclei and several cortical levels for the visual, auditory, somatosensory, and motor systems, are highly similar across primates, and thus results from marmosets are relevant for making inferences about how these systems are organized and function in humans. Nevertheless, the structures in these systems are not identical across primate species, and homologous structures are much bigger and therefore function somewhat differently in human brains. In particular, the large human brain has more cortical areas that add to the complexity of information processing and storage, as well as decision-making, while making new abilities possible, such as language. Thus, inferences about human brains based on studies on marmoset brains alone should be made with a bit of caution.

INTRODUCTION

Marmosets have recently become a very popular research model, largely due to their potential for study with optogenetic and related methods that are now applied to the brains of mice.¹ Marmosets are also of special interest because they are highly vocal and social.² Marmosets provide the research advantages of being a small, rapidly maturing and reproducing primate with a relatively smooth cortex of few fissures.³ Marmosets and the closely related tamarins comprise a total of 20 or more species in South and Central America that vary in appearances and size. Overall, they are the smallest of monkeys, with the pygmy marmoset weighing just over 100 g as an adult. As a small, New World monkey, their small brains still share many features with those of more studied primates such as New World squirrel monkeys and Old World macaque monkeys. Marmosets are thought to have evolved from larger ancestors.^4,5 As they became smaller, marmosets were more able to feed in the thin outer branches of bushes and trees and cling to vertical trunks, giving them access to food less available to larger monkeys.⁶ As larger ancestors evolved into marmosets, their smaller brains likely became somewhat simplified, as small brains in small mammals face the dilemma of reducing the sizes and functions of cortical areas and subcortical nuclei overall, or reducing the numbers of distinct areas and nuclei to save space for the more critical cortical structures.⁷ There is some evidence that mammals, and especially primates, with larger brains have more cortical areas,⁸ and it seems rather certain that the large human brains have more functional subdivisions than most or all other primates, especially marmosets. Thus, marmoset brains can be studied for basic primate features, but the brains of other studied primates, such as macaques, may more closely resemble human brains.^9,10 Other notable specializations of marmosets include having clawed fingers and toes for clinging to trees and generally giving birth to 2 offspring. The common marmoset, Callithrix jacchus, is the marmoset most used in research. The focus here is on the features of the marmoset forebrain, especially those in neocortex, compared with other primates. Fukushima et al¹¹ provided an excellent recent review of marmoset neuroanatomy.

DEFINING ARCHITECTONIC AREAS AND NUCLEI

Defining the nuclei of the brainstem and thalamus and areas of neocortex is at the center of understanding how structures and networks in the nervous system process information, store memories, and plan and create motor responses. Defining these subdivisions of brains is basic to further advances that include understandings of their functions and roles in behavior. In this regard, researchers are fortunate in that the recently published stereotaxic atlas of the marmoset brain¹² is one the best of modern efforts, and one that certainly exceeds the classical descriptions of Brodmann.¹³ While the atlas defines subcortical structures in ways that do not differ very much from earlier descriptions in monkeys and other primates,¹⁴ the Paxinos et al¹² atlas recognizes a total of 116 histochemically defined cortical areas per hemisphere in marmosets, one that approaches the currently defined estimates for the much larger brains of macaque (~140)¹⁵ and humans (~180).¹⁶ In contrast, the classical description of cortical areas by Brodmann¹³ (1909) only included about 30 areas for marmosets and just 47 for humans. There are reasons for the larger numbers of areas now proposed for taxa of primates. First, there are now many more histological procedures for revealing structural differences that are used to histologically define areas and nuclei. Brodmann¹³ depended on the Nissl stain for revealing structural differences between proposed areas based on cell bodies, whereas the recent Paxinos et al¹² atlas was based on coronal brain sections stained for Nissl substance, calbindin, acetylcholinesterase, and non-phosphorylated neurofilaments (SMI-32). In addition, there are many other histochemical and immunostaining procedures available,¹⁷ with each revealing somewhat different aspects of regional differences in brain structure. Second, and more importantly, modern atlases are guided by information from studies on the connections and functional maps of cortical and subcortical structures. For example, the early architectonic study of the cortical areas of a mouse lemur brain by Le Gros Clark in 1931 did not define an architectonically distinct area now found in all primates,¹⁸ the middle temporal visual area (MT),¹⁹ but the MT was included in later architectonic studies of mouse lemur cortex.^20,21 Modern depictions of cortical areas in primates include an MT (Fig. 1), which is easily identified architectonically by multiple procedures.

Figure 1 .

A dorsolateral view of a marmoset brain with proposed functional subdivisions. Visual areas have been named as follows. The primary or first visual area is V1 or area 17. Other commonly named visual areas include the second (V2), third (V3), and both (V4) visual areas. Alternatively, V4 is called the dorsolateral visual area (DL). DL appears to have rostral (DLr) and caudal (DLc) divisions. The nature of V3 in marmosets and other primates is uncertain. Here, V3 is disrupted by part of the dorsomedial visual area (DM) and the medial visual area (M) on the medial wall of the cerebral hemisphere. The middle temporal visual area (MT) is surrounded by functionally related visual areas: the MT crescent (MTc), the middle superior temporal area (MST), and the dorsal (FSTd) and ventral (FSTv) fundal superior temporal areas. The caudal temporal lobe is mainly visual, and 2 traditional divisions are illustrated, the temporal region (TE), and temporal-occipital region (TEO). Additionally, 3 of the regions activated by seeing faces (F) are outlined. Auditory cortex includes 3 primary areas, A1, the rostral area (R) and the rostral temporal area (RT). The second and third level of cortical processing include subdivisions of the auditory belt and the auditory parabelt (see Fig. 2). Other parts of the auditory temporal lobe are included in further processing of auditory information (Aud). Somatosensory cortex includes 4 somatotopically organized areas named after the numbering system of Brodmann,¹³ area 3b or the primary area, S1, and areas 3a, 1 and 2. The existence of area 2 in marmosets is not established. Other areas include the second somatosensory area (S2), the parietal ventral area (PV), and areas in the cortex of the lateral fissure (see Fig. 3). Motor areas include primary motor cortex (M1), the ventral (PMV) and dorsal (PMD) premotor areas, the supplementary motor area (SMA), the frontal eye field (FEF), and the related frontal visual area (FV). The frontal pole of the cerebral hemisphere is occupied by area 10 of Brodmann (09), just over the olfactory bulb. Posterior parietal cortex (PPC) includes rostral (R) PPC that is subdivided into a number of action specific domains with somatosensory and visual inputs and caudal (C) PPC with visual inputs. In area 3b (S1), the hand region is indicated (H) and 4 modules, representing teeth, tongue, teeth, and tongue, are outlined (compare with Fig. 4).

The point here is that more accurate depictions of areas and nuclei will always be guided by additional information and criteria, as areas and nuclei should be identified not only by their histological appearance but by connections with other structures, functional organization such as having a systematic representation of a sensory surface, and neuron response properties.²² The goal is to define the same structures (eg, homologous structures) in different taxa by shared characteristics while recognizing that shared structures have continued to evolve independently in different lines of descent so that they come to differ in many ways while still coming from a common ancestor. Overall, identifying homologous structures despite their acquired differences is not an easy task. By definition, homologous structures are those that were also present in a common ancestor of the species under consideration. But we do not have fossil brain structures from potential ancestors, so the inferences of homology are indirect. Thus, present conclusions should often be regarded as works in progress and subject to modification and change. This will become more apparent as the organizations of anatomical systems in the marmoset are outlined here. Fortunately, there have been many helpful studies of anatomical connections and sensory and motor maps in marmosets.

One of the most distinctive architectonic features of neocortex is that neurons vary in size across cortical layers and cortical areas. In general, cortical areas toward the front of the brain tend to have larger and less densely packed neurons on average than areas toward the back of the brain.²³ In addition, primate brains have more densely packed neurons and therefore smaller neurons on average than rodents, carnivores, marsupials, and other mammals.²⁴ Furthermore, the change in average neuron size from frontal to occipital cortex is much more pronounced in primates than other studied mammals. Marmosets fit nicely into this pattern.²⁵ One possible explanation for the frontal-caudal gradients in cortex is that developmental programs in mammals of “late makes great”^23,26 result in later maturing posterior regions of cortex having more neurons. An alternative explanation considers the functions of cortical areas and proposes that homologous areas across taxa may differ as they specialize in different ways. Considering only frontal-caudal gradient in average neuron size misses something that an area-by-area analysis reveals. Sensory areas tend to preserve the information coming in from the nuclei in the thalamus, and they therefore have small neurons with small dendritic arbors that are responsive to only a few thalamic inputs. This emphasis is most pronounced in layer 4 where neurons are activated by thalamic inputs, but pyramidal neurons with outputs to other cortical areas also tend to be smaller in sensory areas.²⁷ Thus, neurons in primary sensory areas such as primary auditory, somatosensory, and especially visual cortex have the smallest and most densely packed neurons overall, especially in primates,²⁸ including marmosets.²⁵ Motor areas have little or no layer 4 and tend to have large pyramidal neurons that receive many inputs, with those in layer 5 projecting to distant subcortical targets. Thus, they have the largest neurons overall. Considerations of frontal-caudal gradients of neuron sizes and packing densities across cortex thereby overlook important areal differences in neuron packing densities that are more pronounced in some species than others.

The most common other model of the primate brain is that of the macaque monkey. Clear differences in many features, including brain size and numbers of neurons, suggest that different models best address different questions about human brain organization and functions. Marmosets are roughly 10 times smaller than macaques and have less than one-tenth the brain mass and one-tenth the number of neurons.²⁴ The most comprehensive study of neuronal packing densities across cortical areas and regions is that of Atapour et al.²⁵ As in other primates, the primary visual cortex (V1) of marmosets has the most densely packed neurons and thus the smallest neurons on average. Of course, the actual numbers depend on where in V1 the sample for counting is taken, as more neurons across the depth of the cortex occur in the part of V1 representing central vision compared with peripheral vision.

The Visual System: Subcortical Structures

A good place to start describing the marmoset brain is with the visual system as this system has been most extensively studied in marmosets and other primates. The projections of the retina have long been used to identify homologous visual structures in primates and other mammals. In particular, the major targets include the accessory optic nuclei, a part of the pulvinar complex (the lateral posterior “nucleus” of non-primate mammals), the pretectal nuclei, the ventral lateral geniculate nucleus, the dorsal lateral geniculate nucleus, and the superior colliculus (for identification, see the Paxinos et al atlas¹²). Here, the focus is on the dorsal lateral geniculate nucleus (LGN), the visual pulvinar, and the superior colliculus (SC). The retinal projections to the LGN have been frequently studied in part because the laminar pattern of neuron types and the distributions of inputs from the contralateral and ipsilateral eyes are highly variable across taxa.²⁹ Early studies of the projections of the retina to the LGN of primates revealed major differences between the lamination patterns of strepsirrhine and platyrrhine (anthropoid) primates.³⁰ Strepsirrhine primates were formerly known as prosimians, but this classification included tarsiers, which are now known to be an early branch of the anthropoid radiation.

The basic laminar pattern of the anthropoid LGN is of 2 ventral layers of large neurons (the magnocellular, or M layers) with inputs from the contralateral or ipsilateral eye, and 2 dorsal layers of smaller neurons (the parvocellular, or P layers) with inputs from the contralateral or ipsilateral eye, as shown for marmosets (Fig. 2),^31,32 and a scattering of very small koniocellular, or K, neurons between layers (see Huo et al³³ for marmosets). In large primates with proportionately more parvocellular neurons, the parvocellular layers partially subdivide and interdigitate to provide a more complex laminar pattern of parvocellular sublayers. The LGNs of strepsirrhine primates differ by having 2 distinct layers of koniocellular neurons between the 2 dorsal parvocellular layers. The 3 morphological and functional classes of LGN neurons get inputs from 3 matching classes of retinal ganglion cells,³⁴ with roughly 80% of the retinal ganglion cells of primates projecting only to the parvocellular layers, while those projecting to the M or K layers also project to the SC.^35,36 All LGN layers project to the primary visual cortex, V1, or area 17, with the P cell inputs being most important for object vision and the M cell for detecting motion and change.³⁷ The K cells appear to have varied roles, but they were apparently important in early primates that were basically nocturnal. K cells also project to several visual areas in addition to V1. All of the LGN layers are retinotopically organized from central or foveal vision to the contralateral peripheral temporal vision. The layers with inputs from the contralateral eye include the representation of the extreme temporal vision, the monocular visual field.

Figure 2 .

The lateral geniculate of a marmoset. The drawing is based on a coronal brain section with medial to the left. There are 4 obvious geniculate layers, including a pair of more dorsal parvocellular (M) layers and a more ventral pair of magnocellular (M) layers. The outer MC and PC layers receive retinal inputs from the contralateral retina, while the inner MI and PI layers receive inputs from the ipsilateral retina. Small koniocellular (K) neurons exist in the cell poor septal regions between layers, with a larger number between the inner M and P layers and below the MC layer. These septal regions of K neurons are not recognized as distinct layers.

In early vertebrates, the optic tectum (the SC) was the major target of the retinal projections, and the SC remains an important target in present-day mammals. In primates, the SC receives inputs from the M and K pathways that also project to the LGN, but not from the P pathway. The SC of primates represents the complete contralateral hemifields of both eyes but not the complete contralateral retina, as in other mammals. The retinal projections from the contralateral eye terminate most superficially in the superficial grey of the SC in marmosets, as in other primates, with the inputs from the ipsilateral eye terminating just below those of the contralateral eye.³⁵ Other visual inputs from visual areas of cortex terminate deeper, as do somatosensory, auditory, and multisensory inputs from various sources.³⁸ As in other primates, many cortical areas project to the SC in marmosets, with the densest projections coming from 3 visual areas: V1, V2, and MT.³⁹ Outputs of the SC include those to the brainstem for evoking eye movements and those to the LGN and pulvinar nuclei.

The pulvinar of primates is a complex of several nuclei that are mostly visual in function.⁴⁰ The traditional inferior pulvinar (PI) has 4 nuclei: the posterior (PIp), medial (PIm), central medial (PIcm), and central lateral. The traditional lateral pulvinar (PL) has 1 large nucleus (PL or PLvl) and 1 or 2 smaller nuclei. The traditional medial pulvinar (PM) is multisensory and may have functional subdivisions. These nuclei can be distinguished by their different histological characteristics and by their connections. While there has been some uncertainty about where the SC projections to the pulvinar terminate, results from marmosets were among the first to indicate that the SC projects densely to PIp and PIcm and not to PIm.⁴¹ PIm receives inputs from cortical area MT and projects back to MT.⁴⁰ In marmosets, and probably in all other primates, the PIm gets inputs directly from the retina.^31,42 PIp and PIcm get information about visual motion from the SC, and they project to visual areas around MT (areas MT crescent [MTc], medial superior temporal area [MST], FST, and PIcl (the central lateral nucleus of the inferior pulvinar) and PL have modulating interconnections with V1 and other visual areas.^33,40,43 PM has connections with association areas of cortex and has roles in cognitive functions.⁴⁴

Visual Cortex: Primary and Secondary Areas (V1 and V2) and Areas of the Third Tier

The organization of visual cortex in marmosets and other primates, including primary visual cortex and a number of adjoining or nearby visual areas (Fig. 1), is not completely settled. However, the primary area, V1 or area 17, is so obvious architectonically that there is no disagreement on its location and boundaries. Likewise, the MT is so distinct architectonically that there is general agreement about MT’s location and boundaries. The second area, V2 or area 18, has been well studied and is firmly established as a visual area. The main concerns are over the territory of visual area V3 and even its existence as well as the areal extent and organization of the “fourth” visual area, V4, also known as the dorsolateral visual area (DL).

The primary visual cortex is the largest area of the marmoset brain, occupying about 20% of the surface of neocortex.⁴⁵ This is a larger proportion than in primates with larger brains, and it fits the premise that primary sensory areas occupy proportionately more of the cortex in small brains at the expense of higher order areas.⁴⁶ Only the small and visually specialized tarsier has more cortex occupied by V1.⁴⁷ The retinotopy of V1 has been mapped with microelectrode recordings in marmosets, and central vision has an expanded representation in V1,^48,49 as in other diurnal anthropoid primates. The layers and sublayers of V1 (area 17) are well-differentiated in marmosets,^50,51 as they are in other anthropoid primates, but the proposed homologs of these layers with other mammals, including strepsirrhine primates, has been reconsidered.³⁷ V1 or area 17 is known across mammals for having a middle layer 4 of smaller, more densely packed neurons.

Brodmann¹³ described area 17 in a number of mammalian species and came to the conclusion that anthropoid primates had an expanded layer 4 with 2 deeper sublayers, 4Cα, and 4Cβ, of small, densely packed neurons separated from a narrow, more superficial thin layer of moderately small neurons, sublayer 4A, by a sublayer 4B of less densely packed larger neurons. In a more comprehensive review of the laminar pattern of area 17 in primates, Hassler⁵² revised Brodmann’s laminar pattern and concluded that only sublayers 4Cα and 4Cβ correspond to layer 4 of area 17 in other mammals and that sublayers 4A and 4B are actually sublayers of layer 3. This revision of the laminar pattern is strongly supported by more recent evidence that Brodmann’s layer 4B has neurons that project to area V2 and area MT of extrastriate cortex in marmosets^53,54 and other primates.³⁷ Sublayers of layer 4 are not expected to project to other cortical areas, whereas sublayers of layer 3 do project to other visual areas. Sublayer 4A of Brodmann in primates, however, is unusual in that this sublayer has smaller neurons and gets projections from the koniocellular LGN neurons.³⁷ Layer 4 proper in marmosets⁵¹ and other primates receives inputs from the parvocellular and magnocellular LGN layers, with the magnocellular inputs superficial to the parvocellular inputs.³⁷ Layer 6 projects back to the LGN and receives some LGN inputs; both layers 6 and 5 project to the pulvinar, while layer 5 projects to the SC.⁵⁵

Area 17 is an area that has interhemispheric connections between the 2 hemispheres only along its outer border with V2. These connections extend further into V1 in marmosets than in macaques but not as far as in nocturnal primates.⁵⁶ Area 17 also has other features that are shared with other primates, including the layer 3 (layer 4A of Brodmann) “blobs” of high cytochrome oxidase (CO) activity that receive inputs from the koniocellular layers of the LGN.³⁷ The band-like ocular dominance columns that are related to the contralateral or ipsilateral eyes in macaques and humans are only weakly segregated from each other in marmosets.⁵⁷

The second visual area, V2, is an area common to almost all mammals.⁵⁸ An exception is the blind mole rat, which only has an architectonic remnant of area 17. Strangely, V2 is not always recognized in rats and mice by most current investigators. V2 borders V1 along adjoining representation of the zero-vertical meridian through the center of gaze. V2 contains a complete representation of the contralateral visual hemifield, with the outer border of V2 representing the zero-horizontal meridian.⁵⁹ V2 has been mapped in detail by microelectrode recordings in marmosets.⁶⁰ V2 has a prominent layer 4 and is densely packed with neurons but less so than in layer 4 of V1.²⁵ In brain sections cut parallel to the brain surface and stained for CO, the width of V2 in marmosets is crossed by CO dark bands alternating with CO light bands,^61,62 as in other anthropoid primates. These CO dark bands have been characterized as alternating between thick and thin types in macaques,⁶³ leading to the conclusion, based on connections, that there are sets of 4 functionally distinct bands that repeat along the length of V2 in marmosets⁶⁴ and other primates. Importantly, the thick CO bands in V2 relay magnocellular stream information from layer 3C neurons in V1 to MT.^54,55 V2 also projects to V3 and DM.⁶² Subcortical projections of V1, V2, and V3 in marmosets and other monkeys include those to the SC³⁹ and nuclei of the pulvinar.⁶⁵

Area V3 is a cortical field that was proposed on the basis of demonstrated connections of dorsal V1 representing the lower visual hemifield to adjoining dorsal V2 and cortex just rostral to V2 in macaque monkeys.⁶⁶ V3 was proposed as having a mirror image retinotopy of V2, reversing along a common representation of the zero-horizontal meridian at the V2-V3 border. Microelectrode mapping of the retinotopy of the V3 region in macaques subsequently provided compelling evidence for the proposed organization of V3.⁶⁷ However, an early problem for this interpretation was that 2 locations along the rostral border of V2 did not represent the lower visual quadrant as expected for V3, but they represented instead locations in the upper visual quadrant, thus providing evidence for an upper visual quadrant representation of 2 proposed visual areas along the rostral border of dorsal V2, DM, and the medial area (M).^68,69 Although ventral V3 represents the upper visual quadrant as expected for V3, researchers thought that the ventral V3 region, renamed the ventral posterior visual area (VP), did not receive inputs from V1, while clear evidence existed for such a projection to dorsal V3.⁷⁰

These problems for the traditional interpretation of V3 were addressed in 2 different ways by researchers working with marmosets. As the problem of VP as a visual area is that it represented only the upper visual quadrant, researchers combined the lateral part of dorsal V3 with VP to create a new ventrolateral (VL) visual area with at least a partial representation of the lower visual quadrant. Another part of dorsal V3 remained as part of DM.^71,72 This is the interpretation portrayed in the Paxinos et al¹² atlas. Another interpretation followed a study of the projections of both dorsal and ventral V1 in marmosets⁶² and other primates.^73,74 Injections of tracers were placed in different parts of ventral and dorsal V1, and dorsal or ventral parts of V3 were always labeled in expected locations just rostral to V2. This argues that dorsal and ventral parts of V3 combine to form a single retinotopic area just rostral to V2, as previously proposed for V3. However, this interpretation does not eliminate the possibility that dorsal V3 has 1 or 2 gaps that represent the upper visual field of other smaller areas. Such a gap has been mentioned as occurring in some cases of mapping V3 in macaque monkeys, and clear evidence for such a gap was obtained in optical image recordings from V3 in a strepsirrhine (prosimian) primate, the African galago.⁷⁵ Thus, the interpretation of a V3 with 1 or more gaps, perhaps for the upper field representations of DM and M, seems compatible with more of the data.⁷⁶ What is not explained is why dorsal V3, and not ventral V3, has gaps.

Visual Cortex: The Dorsal and Ventral Streams

In marmosets,⁶² and other primates,^73,74 direct, topographically organized projections from V1 provide evidence for 4 retinotopically organized visual areas. One of those targets, V2, is broadly accepted as a valid visual area across primates and in other mammals.⁵⁸ Another is the MT,¹⁹ which appears to exist only in primates. Others include visual area V3, which is open to other interpretations (see above), and the DM (Fig. 4). M gets little or no direct input from V1 and is known mainly from microelectrode mapping data.^69,77 In marmosets, the part of area M (Fig. 4) that represents the upper visual quadrant along the border with V1 has also been called area 19 M.³ This region has been variously divided and named in macaques.^78,79 Thus, there are uncertainties about the organization and name for the visual region of M. Area DM was first described in owl monkeys as a complete visual area representing both upper and lower visual quadrants, and having more myelination than adjoined cortex, in owl monkeys.⁶⁸ The region of DM has been called DM in other primates, including macaques.⁸⁰ Overall, the DM region has been somewhat differently described. For example, Rosa et al⁸¹ and Lyon and Kaas⁶² have different depictions in marmosets. However, it is clear that part of the rostral border of V2 in primates is joined by the cortex representing the upper visual cortex as originally portrayed,⁶⁸ likely occurring in the “gap” in dorsal V3 that represents the upper visual quadrant,⁷⁵ but whether cortex medial to this gap is part of dorsal V3 or the representation of the lower visual quadrant of DM is unsettled.

Figure 4 .

The somatosensory thalamus and cortex of marmosets. (A) The somatosensory thalamus includes 3 main nuclei, the ventroposterior nucleus (VP) which is divided into a lateral division (VPL) that represents the lower body, and a medial division (VPM) that represents the face and mouth. Cell-poor septa separate the hand representation in VPL from that of the face in VPM, as well as the hand and the foot representations in VPL. The more medial parts of VPM represent the teeth and tongue. Most medially, the ventroposterior medial parvocellular nucleus (VPMpc) represents taste. (B) Somatosensory areas and primary motor cortex (M1). Primary somatosensory cortex, S1 or area 3b, represents the contralateral body from tail and foot to hand (H), followed by 3 modules representing different parts of the face (F), and more ventrally by 4 modules representing the contralateral teeth (Te) and tongue (Ton) followed by the ipsilateral teeth and tongue. Area 3a, just rostral to 3b, represents proprioceptors, and parts of 3a may intrude into 3b. Area 1 contains a mirror image reversal of the somatotopy of 3b, and it appears to continue along representations of the face, teeth, and tongue in area 3b. In area 2, a representation of both touch and proprioception is expected, but there is only limited evidence for area 2 in marmosets. The second somatosensory area (S2) adjoins area 3b or an extension of area 1 and extends into the lateral sulcus. Another somatosensory representation, the parietal ventral area (PV) adjoins S2 ventrally. Deeper in the lateral sulcus, the ventral somatosensory region is divided into caudal (VSc) and rostral (VSr) areas. Just rostral to area 3a, the primary motor area (M1) contains crude somatotopic representation from leg to mouth, but M1 is more complexly organized into action-specific modules that may involve more than 1 body part (arm and hand, arm and face, arm and legs) when compared with Figure 1.

There is more agreement over the organization of the region of visual area MT in primates, including marmosets. MT is an easy area to define by its marked myelination and dense expression of CO. MT was first described in owl monkeys by clear histological characteristics, a representation of the visual hemifield, and neurons responsive to stimulus movement and stimulus onset.¹⁹ Major activating inputs come from both stellate and pyramidal neurons in V1.⁸² Optical imaging of evoked activity reveals that MT contains systematic maps of stimulus orientation and direction of motion responses relative to retinotopy.⁸³ Studies on marmosets have the advantage of having MT completely located on the surface of cortex due to the short temporal sulcus, and MT and surrounding visual areas have been mapped with microelectrodes by Rosa and Elston in 1998.⁶⁰ In addition, the cortical connections of MT with MST, FSTd, MTc (Fig. 1), and other regions have been described in marmosets.^53,61,62 Subcortical inputs to MT include those from the medial nucleus (PIm) of inferior pulvinar, as well as from the koniocellular cells of the LGN.⁴¹ These connections are similar to those reported for other monkeys.⁶¹ Because it is easy to identify, MT has been found in all studied primates.

Studies of the postnatal development of MT in marmosets suggest that MT develops early, in part because of direct retinal inputs to the PIm of the pulvinar that relay to MT. Thereby, MT serves as an early organizer of surrounding visual areas and directs the subsequent development of visual-motor areas in the posterior parietal cortex.^40,42,84 Much of MT’s influence on higher order visual processing comes from surrounding areas with short, direct inputs from MT. These areas include the MST, just rostral to MT; a ventral area, FST, which has been divided in dorsal (FSTd) and ventral (FSTv) areas with direct (FSTd) and indirect (FSTv) connections with MT; and a nearly surrounding band-like area, the MTc.⁸⁵ MTc is especially distinct as a crescent of CO ovals in a CO light band around MT.⁶² All 3 areas have separate representations of the contralateral visual hemifield, and they have been studied in marmosets.⁵³ These areas, together with MT and DM, are major contributors to visual and visuomotor areas of posterior parietal cortex that form the extensions of the dorsal stream of visual processing into parietal and frontal cortex.^40,86

Visual areas of the ventral stream of visual processing in the temporal lobe are less well understood in marmosets. According to current views of the ventral stream of visual processing, which mediates object vision in primates,⁸⁷ areas V1, V2, and V3 contribute to both dorsal and ventral streams via segregated pathways through these 3 areas.⁶⁴ The cortical pathways with the information essential for the ventral stream terminate in the dorsolateral area DL (V4) located between central V3 and MT (Fig. 1). DL was originally described as a retinotopic visual area in owl monkeys located between V2 and MT.^59,88 The original DL included the region of MTc and likely parts of V3. The main part of DL represented central vision, and the proposed extensions of DL along MT in what is now considered to be MTc were included because they added paracentral and peripheral vision. The DL region was called V4 in macaques, based on topographic connections with V1, V2, and V3,⁸⁹ and V4 is the term most commonly used today. The V4-DL region in macaques appears to have 2 separate parallel representations within rostral and caudal bands, DLr and DLc, as the connections of these 2 regions differ, even after MTc is removed from V4/DL.⁹⁰ Connection patterns with especially V2 have been used to delineate the dorsal border or DL/V4, but the location of the ventral border is still somewhat uncertain.⁹⁰ The Paxinos et al¹² 2012 atlas of marmoset cortex illustrates a single area as V4 or the ventrolateral area (VL), but this V4 extends so far ventrally into the temporal lobe that much of this V4 or VLA area would represent the upper visual quadrant, making it much larger than the proposed more dorsal representation of the lower visual quadrant. While DLr appears to have connections more related to the dorsal stream of processing, DLc provides the major direct inputs to the upper temporal lobe (area TEO) as the next step in processing in the ventral stream. Further processing in the ventral stream involves areas and regions that have been defined functionally as critical in object vision, but they do not correspond to the proposed subdivisions of the temporal lobe based on classical cytoarchitecture.

In humans and macaques, parts of the lower temporal lobe are specialized for the processing of faces because faces are important in social signaling.^91–93 More recently, such face-sensitive patches (Fig. 1) have been shown in the caudoventral temporal cortex of marmosets,⁹⁴ suggesting that the face-processing network evolved in early monkeys at least 35 million years ago and that social visual perception relevant to humans can be studied in marmosets. See Kravitz et al⁹⁵ for a review of the ventral visual pathway.

THE AUDITORY SYSTEM

Early mammals differed from their reptilian-like ancestors by modifying the mechanics of sound conduction to the inner ear so that hearing of higher frequencies was possible. Some species of bats are now known to have further adaptations that allow them to navigate and locate flying insects with very high frequency sounds. Primates do not have these extensive peripheral adaptations, but they do have specializations for vocal communications between members of the same species.⁹⁶ The auditory system of marmosets has been studied, in part, because they are very social primates that communicate extensively with a variety of calls.^97,98 As these studies have focused on higher levels of auditory processing, cortical processing is emphasized here.

Across mammals, the external ear and ear canal conduct sound pressure waves to the eardrum, where these vibrations are conducted by 3 bones to the oval window of the cochlea, a long-coiled tube of 3 compartments surrounded by bone. The middle compartment has inner sensory hair cells that respond to vibrations from sounds in the cochlear fluids, and these inner hair cells activate the afferents of ganglion cells that have axons that form the auditory nerve. The hair cells and axons of the auditory nerve encode information on the frequencies and intensities of sounds. Compared with the somatosensory and visual systems, more of the processing of sensory inputs are occurs subcortically in a complex of brainstem structures that are similar across mammalian taxa.⁹⁹ The auditory nerve projects to divisions of the cochlear complex, the dorsal, posterior, and anterior cochlear nuclei that project contralaterally to brainstem structures and bilaterally to the superior olivary complex. The contralateral projections also go to the nuclei of the ascending pathway, the lateral lemniscus, and to the inferior colliculus. Other ascending projections include those from the superior olivary complex. The connections between the nuclei of the lateral lemniscus and the inferior colliculus of both sides further allow neurons to compare outputs from the 2 ears for information on sound location. The inferior colliculus receives inputs from cochlear and other subcortical nuclei as well as from descending inputs from the SC, thalamus, and cortex. The inferior colliculus contains at least 3 functionally distinct subdivisions.^100,101 The large laminated central nucleus consists of sheets of neurons that systematically represent sound frequencies with low frequencies represented in sheets dorsolateral to those for high frequencies.¹⁰² The central nucleus of the inferior colliculus projects topographically to the tonotopically organized ventral nucleus of the medial geniculate complex of the auditory thalamus; while these projections are bilateral, the ipsilateral projections are much denser. The dorsal cortex and the external lateral nucleus of the inferior colliculus project to dorsal and medial nuclei of the medial geniculate complex of the auditory thalamus.¹⁰³ As in other primates, the projections from auditory cortex to the inferior colliculus in marmosets are most dense in the dorsal cortex, but sparse in the central nucleus.¹⁰⁴

The auditory thalamus includes 3 traditional architectonic divisions of the medial geniculate complex: the ventral (MGv), dorsal (MGd), and medial or magnocellular (MGm) nuclei.¹⁴ These nuclei have been identified in many species of mammals, including several primate species. In marmosets, these and other related nuclei have been fully described by de la Mothe et al¹⁰⁵ and Bartlett and Wang.¹⁰⁶ The large MGv has tightly packed neurons of medium size and most densely express CO and the glutamate transporter, VGLUT2. MGv receives tonotopically organized projections from the central nucleus of the ipsilateral inferior colliculus but also sparse projections from the contralateral central nucleus. The projections terminate in thin layers that successively represent low-to-high-tone frequencies in a ventrocaudal-to-dorsorostral progression.¹⁰⁷ Relay neurons in MGv project densely to layer 4 of the primary areas of auditory cortex, where they are thought to be the major activating source of input. MGv has only sparse projections to secondary areas of the auditory belt.^105,108 MGv has densely packed, smaller neurons.¹⁰⁴ MGd receives inputs from the dorsal cortex of the inferior colliculus. Neurons in MGd are broadly responsive to tone frequencies and may respond best to complex sounds. Projections of MGd are to areas at the second and third levels of cortical processing outside the auditory core. MGm has neurons of several sizes, including the large magnocellular neurons.¹⁰⁸ Neurons in MGm vary in responsiveness to tone and other stimuli, and some have very short response latencies.¹⁰⁹ Both MGd and MGm project mainly to non-primary areas of auditory cortex, and where they terminate in superficial layers 2 and 3, with probably a modulatory role.^104,105,107

Several other thalamic nuclei have roles in auditory processing, including the suprageniculate nucleus with multisensory inputs from the SC and broad connections to auditory and multisensory cortex.¹⁰⁸ The PM projects broadly to multisensory areas of cortex, including the third cortical level of auditory processing, the auditory parabelt.

Auditory Cortex

Current theories for how auditory cortex are organized in primates came out of a series of experiments on macaque monkeys,^{108,110–113} owl monkeys,^107,114 and marmosets.^{105,115–117} The cortical model that emerged (Figs. 1 and 3) appears to apply broadly to anthropoid primates, including apes and humans.¹¹⁸ This model of auditory cortex includes a core of 3 primary (or primary-like) areas, the traditional primary area (A1), the more recently defined rostral auditory area (R), and the rostrotemporal area (RT). Surrounding the core, a belt of secondary areas, named by location relative to the core, includes the caudolateral area, middle lateral area, and the rostrotemporal lateral area of the lateral belt and the caudolateral area, rostromedial area, and the rostrotemporal area of the medial belt. The parabelt region includes 2 divisions: the rostral parabelt and the caudal parabelt. The core is distinguished from the belt architectonically, and areas within the core are distinguished by patterns of tonotopic organization. The belt areas receive inputs directly from the core areas, and the areas of the belt are identified by their different patterns of connections. The parabelt region has dense inputs from the belt areas, and the 2 divisions of the parabelt are distinguished by their different patterns of input from the belt areas. Further study might reveal more divisions of the parabelt and other modifications of the cortical model, but it now is well supported by a large body of evidence.

Figure 3 .

A surface view of flattened auditory cortex of marmosets. The region is divided into a ventral part that is visible on the exposed surface of the temporal lobe and a part unfolded from the caudal bank of the lateral sulcus (LS). Here both parts are shown on a flattened surface view of auditory cortex. The part of the lateral sulcus that divides auditory cortex is marked by a dashed line. The 3 primary areas are the traditional primary area (A1), the rostral area (R), and the rostral temporal area (RT). These areas represent high (H) to low (L) tone frequencies in progressions that reverse at areal borders. The auditory belt includes the rostral temporal lateral (RTL), the anterior lateral (AL), the medial lateral, and the caudal lateral (CL) areas of the lateral belt, and the rostral temporal medial (RTM), rostral medial (RM), and caudal rostral (CR) areas of the medial belt. The third level of auditory processing of the parabelt includes rostral (RPB) and caudal (CPB) parabelt divisions (compare with Fig. 1).

The primary area (A1) was the first area explored in microelectrode recording studies in primates.¹¹³ The general assumption was that mammals had only 1 primary area, as proposed for the visual and somatosensory systems.¹¹⁹ Through microelectrode mapping studies of tonotopic patterns, it became clear that rodents, cats, primates, and likely other mammals had more than 1 area with primary-like characteristics. The problem was determining which one of the primary-like areas was the primary area A1. As the primary-like areas in cats were described, A1 became the tonotopic area that occupied much of the territory of the originally described A1 with less precise methods. Yet it became clear that the adjoining anterior auditory field and the posterior auditory field in cats were also clearly primary-like in function and anatomy. Rodents also were found to have 3 primary-like fields with an AAF, A1, P progression of tonotopic maps, but some investigators called the AAF map A1. In primates, the core areas of A1, R, and RT do not have obvious homologues with the AAF, A1, P sequences in rodents and cats because the sequences of frequencies in the representations differ. In primates, the A1 name is well established, but it is possible that it is not the same area that is called A1 in other mammals. The existence of 2 or 3 primary-like areas in various mammals suggests that it is advantageous to provide 3 parallel streams at the first cortical stage of auditory processing instead of starting from a single primary area, as for vision and touch. The somewhat different response characteristics of neurons in R, RT, and A1 as well as their cortical connections suggest that R and RT are more associated with a ventral stream at auditory processing for the identification of sounds and communication calls, while A1 is more associated with the dorsal stream of processing for localizing the sources of sounds.^120–122 A1 and adjacent belt and parabelt regions contribute more to dorsal cortical regions that project to dorsal parts of prefrontal cortex, while R and RT and adjacent cortex contribute more to the ventral cortex in the temporal lobe that projects to the ventral prefrontal cortex. In marmosets, RT has projections that include area 10 of prefrontal cortex.¹²³ Overall, the core areas are densely myelinated and distinguished from the belt by expressing more parvalbumin, CO, and acetylcholinesterase. Activating inputs are from MGv. Layer 4 is thicker and more densely packed with smaller neurons in core areas, and neurons respond better to pure tones. One of the belt areas, area CM, is more like a primary area than other belt areas.¹¹⁷ CM has a tonotopic organization, and neurons respond to tones. CM has more of the architectonic characteristics of the core areas than other belt areas, but they are less pronounced than in primary areas. In addition, CM has thalamic inputs from MGd and not MGv, and lesions of A1 abolish the auditory responses of CM neurons.¹²⁴ However, CM may be a key area of the dorsal stream, as neurons in CM are more selective for the locations of sounds.

The outputs of parabelt cortex involve caudally adjacent multisensory cortex, projections to more ventral temporal cortex, inputs to parts of posterior parietal cortex, and projections to frontal cortex.¹²⁵ These connections have been used to identify parts of cortex involved in the dorsal and ventral streams of auditory cortex.^112,126 Both auditory and visual areas project to overlapping regions of the frontal lobe, with a dorsal frontal region more involved with spatial locations and a ventral region more involved in object recognition.^121,122 Some projections from multisensory regions of the temporal lobe in marmosets terminate in the frontal pole region area 10 of neocortex known as in other primates.¹²⁷

THE SOMATOSENSORY SYSTEM

The somatosensory systems of marmosets are overall highly similar to those of other anthropoid primates. Yet, there are likely some specializations related to their small size and to their unique trait of having claws on all fingers and toes except the great toe, unlike other primates. These claws are not a vestigial trait of early primate or non-primate ancestors but a re-evolved trait or specialized feature that allows them to hold on while climbing up and down the vertical trunks of trees, much like squirrels do.⁶ Fingers and toes in other monkeys have nails to protect and reinforce the underlying digit and toe pads, which have high concentrations of tactile receptors. The hands of marmoset are more like paws and thereby less suitable for manipulating objects. Thus, marmosets may obtain proportionally less somatosensory information from their glabrous surface of their hands and feet than other anthropoid primates.

The somatosensory systems of marmosets and other primates include inputs from specialized receptors in the hand and other parts of the body that terminate on neurons in the dorsal horn of the grey matter of the spinal cord or in brainstem nuclei. Tactile inputs from the skin have branches that travel in the dorsal columns of the spinal cord to nuclei that represent the lower body (the gracile nucleus) or the forelimb (cuneate nucleus), or the face and mouth for the principal nucleus of the trigeminal complex in the brainstem. The representations of the hand and arm have been mapped in the cuneate nucleus of marmosets¹²⁸ as well as in other primates. The brainstem nuclei project to the ventroposterior nucleus of the contralateral thalamus, which projects to primary somatosensory cortex, where information is then distributed to higher order somatosensory areas.¹²⁹

The glabrous skin of the hands and feet of primates contain 4 types of low-threshold mechanoreceptors that respond to touch and pressure on the skin. These include the rapidly adapting type 1 that originates in Meissner corpuscles, responds on stimulus onset and offset to taps and flutters, and signals the texture of surface. The rapidly adapting type 1 afferent inputs are complimented by the slowly adapting type 1 afferents that originate in Merkel disks, respond during continued light pressure, and signal edges on surfaces. The less frequent SA2 and RA2 receptors signal skin stretch or vibration. The hairy skin has these types of receptors with some modifications, including touch domes and other receptors sensitive to hair and whisker movements or grooming. The stroking of the fur coat in social species, such as marmosets, may evoke a pleasant sensation known as affiliative touch.¹³⁰ The inputs for affiliative touch and the high-dynamic range neurons, which may contribute to pain, terminate in the dorsal horn of spinal cord or on brainstem neurons that send projections to the contralateral thalamus via the spinothalamic pathways or its equivalent in the brainstem. These afferent pathways and relays to the thalamus have not been much studied in marmosets compared with other primates.

The somatosensory thalamus of primates (Fig. 4A) includes a large ventroposterior nucleus (VP) for inputs from the dorsal column nuclei and the principal nucleus of the trigeminal complex. These inputs are largely those that reflect the rapid and slowly adapting type 1 tactile afferents. VP contains 2 major subdivisions, the ventroposterior lateral subnucleus (VPL), which represents the contralateral body surface from tail to hand, arm, and neck; and the ventroposterior medial subnucleus (VPM), which represents the face, teeth, and tongue. Dorsomedial to VPM, the parvocellular ventroposterior nucleus (VPMpc) represents taste inputs from the tongue. The ventroposterior inferior nucleus (VPI) with inputs from the spinothalamic tract and its equivalent from the trigeminal spinal nucleus of the brainstem is just inferior to VP. The ventroposterior superior nucleus (VPS) just above VP receives proprioceptive inputs from the spinal cord and brainstem.¹²⁹ As in other primates, these nuclei can be defined in marmosets by their architectonic characteristics as well as by inputs from subcortical structures and projections to cortical areas.

In Nissl and CO preparations, VP of marmosets stands out as more darkly stained than surrounding nuclei.^12,131–133 VP is divided into VPL and VPM by a narrow, cell-poor septum that extends dorsally from VPI. Other less apparent septa separate the hand representation from the more lateral foot representation and the 5 digits of the hand from each other. Such septal regions also separate darkly stained clusters of neurons in VPM, such as those for representing the face, tongue, and teeth.¹³¹ the parvocellular ventroposterior nucleus for taste is distinguished from VPM by having somewhat smaller neurons and by being darker in sections stained for calbindin. VPI stains lightly for CO and is sparsely packed with neurons. Krubitzer and Kaas¹³² suggest that the poorly stained septal regions of VP are extensions of VPI into VP. VPS has fewer darkly packed neurons than VPM and less CO. As in other monkeys, VP projects densely to layer 4 of S1 (area 3b) in marmosets.^131,132 The connections are somatotopic with injections of tracer in different parts of the body representation in S1 receiving inputs from different parts of VP. The major thalamic projections to the second somatosensory area (S2) and parietal ventral area (PV) are from VPI and the septal regions of VP.^132,133 The major thalamic inputs to area 3a are from VPS.¹³⁴

The somatosensory cortex in marmosets (Figs. 1 and 4B) includes areas S1 (area 3b), area 1, area 3a, area S2, area PV, and possibly area 2 (Fig. 4B). All these areas have been defined and studied in other primates.¹²⁹ The first limited study of the somatotopic organization of S1 cortex in marmosets or tamarins was reported only in a brief abstract.¹³⁵ Subsequently, a nearly complete representation of the contralateral body surface in tamarins was described in an extensive microelectrode mapping study.¹³⁶ The representation was confined to architectonic area, 3b, and it extended from the foot most medially in precentral parietal cortex (note the lack of a central sulcus in these primates) through the leg, trunk, arm, and wrist to a large representation of the hand with digits 5 to 1 in a mediolateral sequence. More laterally and extending rostrally, representations of the face, teeth, and tongue were recorded, much as in Figure 1. The more ventral second representation of ipsilateral teeth and tongue were not described until much later.¹³¹ Later studies confirmed aspects of this overall somatotopic organization and provided further details in marmosets.^61,137 Additional studies compared the normal organization of the hand in S1 of marmosets with the altered representation after the loss of some afferents from the hand.^138,139 The architectonic features of S1 in marmosets include having densely packed neurons and dense myelination.^12,137 In brain sections cut parallel to the cortical surface, S1 (area 3b) has a pronounced myelin-sparse septum between the hand and face representations^61,140 but not the obvious septal regions between digit representations that can be seen in macaque monkeys.¹⁴¹ As in other monkeys, few intrinsic connections in S1 cross the hand-face border in marmosets.¹⁴⁰

Unlike other monkeys that have been mapped with microelectrodes, there is only limited evidence for a complete representation of the contralateral body surface along the caudal border of S1, the area 1 representation, and only a hint of an area 2 representation.¹³⁸ Likely these areas are present but are not very responsive under anesthesia, as suggested by Carlson et al.¹³⁶ Area 2 representations have been described in macaques and New World cebus monkeys, but an area 2 somatotopic representation has not been reported in any small New World monkey. Thus, the existence of an area 2 has been questioned for marmosets as well as for owl monkeys and titi monkeys.¹⁴² However, there is good evidence for area 3a in marmosets that corresponds to area 3a of other primates as an area with proprioceptive inputs from VPS of the thalamus.^134,137 In addition, there is evidence for the 2 somatosensory representations, the second somatosensory representation (S2) and the PV, in marmosets.^61,133,136 These 2 areas depend on inputs from S1 for activation as lesions of S1 in marmosets render S2 and PV unresponsive.¹⁴³ Other cortical connections of marmosets include sparse connections of S1 with area 3a, connections of 3b with the area 1–2 region, and dense connections between area 3a and area 1–2 region.^134,140

Sensorimotor Cortex

There is experimental evidence that the frontal motor areas of other primates exist in marmosets (Fig. 1). Parts of posterior parietal cortex with parietal–frontal connections also exist, although parietal motor systems have not been extensively studied in marmosets. Importantly, the primary motor area (M1) has been identified and mapped with stimulating microelectrodes, and the M1 of marmosets, as in other primates, represents movements of the contralateral body from hindlimb to face in mediolateral sequences.^137,144,145 M1 is coextensive with an agranular region of cortex that has large pyramidal neurons. Thalamic connections of M1 of marmosets, as expected, are with the ventrolateral nucleus.¹³⁴ Movements have been evoked from area 3a and, at higher levels of current, from parts of S1. Premotor cortex was not very responsive, but limited evidence was obtained for the dorsal and ventral promotor areas (PMD and PMV) as well as the supplementary motor area.¹⁴⁶ The cortical connections of caudal and rostral subdivisions of dorsal premotor cortex have been described in marmosets¹⁴⁷ as similar to those in other monkeys. There is limited evidence from microsimulation and connections from visual areas for a frontal eye field and an adjoining frontal ventral area with visual connections.^61,148 Connections of architectonic fields of prefrontal cortex, including the region of the frontal eye field, have been described in marmosets.¹⁴⁹ Less is known about the organization of sensorimotor areas of posterior parietal cortex that have been well studied in other primates.^86,150 However, single-unit recordings from neurons in posterior parietal cortex during a saccadic eye movement task revealed a region likely to be the lateral intraparietal area of other primates.¹⁵¹ Saccades were evoked from the same region by electrical stimulation.¹⁵² The proposed locations of other sensorimotor areas of posterior parietal cortex have been illustrated by Majka et al.¹⁵³

OTHER CORTICAL AREAS

A description of other proposed architectonic fields and connections of marmosets is beyond the scope of this review. Other regions of cortex are presented in the Paxinos et al¹² atlas and in Majka et al.¹⁵³ Connections of the frontal pole of prefrontal cortex are described by Burman et al.¹²⁷ Prefrontal cortex is important in the recognition of visual and auditory components of vocal communication, and marmosets are highly vocal.¹ Chaplin et al¹⁵⁴ indicate how proportions of regions of cortex have expanded differently in primates with large brains compared with the brains of marmosets. Atapour et al²⁵ provide a detailed description of how the packing densities of neurons in the neocortex vary across cortical areas in marmosets. In addition, proportionally fewer neurons in marmosets are in primary sensory areas of cortex compared with the thalamic relay nuclei in small-brained marmosets than in primates with larger brains.^155,156 This suggests that even for primary sensory areas, primates with larger brains are more dependent on cortical areas, the so called “corticalization of function.” Finally, resting-state functional magnetic resonance imaging has been used in marmosets to reveal large-scale brain networks.¹⁵⁷

Already, marmosets have become one of the most studied of primates in neuroscience, but there is still so much to do.

CONCLUSION

Marmosets are presently considered to be a highly useful model of the primate brain where studies are highly relevant to understandings of the organization and functions of the human brain. As anthropoid primates, marmoset brains share many features with human brains, but human brains are much larger and more complex. Human brains have many more neurons and more functionally distinct subdivisions. Thus, inferences about human brains made from studies on marmoset brains should be made with caution, and information from studies on other model primates, especially from macaque monkeys, should be considered. Nevertheless, there is much to learn from studies on marmoset brains that will be informative and highly relevant. Marmosets offer the advantages of having brains that have already been extensively studied and having a small brain that offers many technical advantages. In addition, they mature and age more rapidly than large primates and thus are useful in studies on development and aging.

CONFLICT OF INTEREST STATEMENT

The author declares no conflicts of interest.