This is the first study we are aware of to globally map movement-associated areas in the avian cerebrum. The discovered areas are adjacent to the cerebral vocal nuclei in all three vocal learning orders. The anatomical extent of the movement-associated areas are larger than the vocal nuclei, which is consistent with a greater amount of musculature involved in the control of limb and body movements relative to that for the syrinx. Below we discuss the implications of our results for understanding motor and somatosensory pathways in birds, and the evolution of brain pathways for vocal learning.
Movement-associated brain areas in birds
It is well established that voluntary movements in mammals are controlled by motor and somatosensory pathways in the cerebrum. Motor cortical areas send commands to lower motor neurons in the brainstem and spinal cord that control muscle contraction and relaxation, and to motor basal ganglia areas that modulate ongoing movements, whereas muscle spindles send proprioceptive feedback to somatosensory areas that sense and modulate ongoing movements 
. However, the motor and somatosensory pathways function in an overlapping manner: the somatosensory cortex sends efference copies of sensory commands to motor cortex prior to the motor command, and the motor cortex has inhibitory connections with the somatosensory cortex to modulate the somatosensory input 
. Although cerebral motor pathways are well understood in the mammalian brain, surprisingly little is known for the avian brain. Here we found that limb and body movements result in activation within specific cerebral areas of the two known somatosensory pathways (AH and AMD; Nb and MVb). There was striking specificity in the activation patterns when other sensory factors (vision and audition) were eliminated, allowing us to map functional domains of the avian cerebrum. Although such specificity cannot be readily revealed by tracer studies, such studies have shown that the AH area of zebra finches, owls, and pigeons receives input from AMD and from the anterior portion of the intercalated lamina of the hyperpallium (IH), which in turn receives input from the somatosensory thalamus and spinal cord dorsal column nuclei, which are innervated by somatosensory neurons from the wings and legs 
. AH also sends descending projections to the intermediate grey matter of the spinal cord 
. Some have interpreted this pyramidal tract-like projection to the spinal cord to indicate that AH may also have motor or mixed somatosensory-motor functions, and is the homolog of the mammalian motor cortex 
. Wild and Williams 
who discovered this projection proposed instead that it is not motor, but somatosensory feedback to the ascending somatosensory pathway of the spinal cord. As for the other pathway, B also receives somatosensory input and projects to Nb 
and Nb projects to MVb 
; it is not clear where MVb projects to. Electrophysiology studies show that B has a somatotopic map of the body in budgerigars 
, whereas AH has a touch somatotopic map of the leg and foot in owls 
. Presumably the areas that sense leg movements during hopping or wing movements during whirring were specifically activated in our study. This activation could conceivably be used for processing proprioceptive feedback from muscles spindles and/or skin touch receptors as the animal sense the floor with its feet or surrounding air with its wings, an idea that can be tested with peripheral stimulation and removal of somatosensory input.
There was also consistent cerebellum activation, as would be expected because the cerebellum receives somatosensory input and sends motor output commands for fine coordination of movements 
. Our findings are the first that we are aware to identify patterns of movement-associated IEG activation in the cerebellum. The cerebellum in birds, as in mammals, has two somatotopic body representations: one in the anterior half from lobules I-VI and the other in posterior lobules IX-X 
; determining more specific topographic organization has yielded conflicting results 
. Connectivity data in pigeons and zebra finches suggest overlapping zones where lobules I-III and IXab receive input from the neck, III-V from the wings, III-VI and IXcd from the legs, VII-VIII from visual and auditory areas but also from somatosensory AH 
. Our findings are partly consistent with this picture, in that wing whirring activated IEG expression mostly in lobules II-VI, hopping in lobules VI and IXcd, and when moving in dim light or in the dark little if any activation occurred in VII-VIII. In general, the cerebellum activation patterns suggest that limb movements may be mostly responsible for the overall brain gene activation seen in the controlled hopping movement groups, as lobules II-VI and IXcd that are connected to the limbs were consistently activated.
With known somatosensory areas identified, a remaining question is where are the motor areas? Besides the hypothesis that AH and AMD are motor in addition to somatosensory 
, previous studies have suggested the arcopallium and dorsal striatum as general motor areas of the avian cerebrum 
. However, none that we are aware of have used movement behavior to map a cerebral motor system. Based on our results, we hypothesize that a general motor system in birds consists of the brain areas adjacent to the cerebral vocal nuclei of vocal learners. Our reasons are as follows: First, we found a close association in locations and size of these movement-associated areas with the cerebral vocal nuclei. Second, like the vocal nuclei 
, these brain areas have movement-associated IEG expression that is independent of auditory and visual input. Third, like the vocal nuclei, the expression levels correlate with the amount of movement performed. Fourth it appears that the vocal nuclei and adjacent regions have similar connectivity, as described below.
Many prior studies have accidentally or purposely placed tracers adjacent to the songbird vocal nuclei, and except for the HVC shelf and RA cup 
the function of these brain regions were not known. In some of these studies on zebra finches, we note remarkable overlap in the connectivity patterns 
with the movement-associated gene expression patterns (this study). When we compile these connectivity results with the movement-associated gene expression results ( 
), it appears that the movement-associated areas in songbirds may be connected in anterior and posterior pathways in parallel, although not identical, with the adjacent vocal nuclei (). The anterior movement-associated areas, like the anterior vocal pathway nuclei, appear to be connected in a pallial-basal-ganglia-thalamic-pallial loop (;
white arrows): anterior AMV to AN adjacent to LMAN, these two areas to the striatum adjacent to Area X, the striatum via its pallidal-like neurons to the dorsal thalamus adjacent to the vocal part of DLM, and the dorsal thalamus back to the AN adjacent to LMAN. Connectivity of MO or the surrounding MV is not known in songbirds, but the comparable song nucleus and adjacent MV in parrots and the MV in pigeons projects to the anterior nidopallial and striatal vocal nuclei and surrounding area, respectively 
. The parrot anterior vocal pathway also forms a pallial-basal-ganglia-thalamic-pallial loop () 
. That is, similar connectivity can be compiled for these cerebral regions in other vocal learning and vocal non-learning birds 
Sources for connectivity of neural populations adjacent to songbird vocal nuclei
Summary of the results of this study and proposed theory.
The posterior movement-associated areas in zebra finches, like the songbird posterior vocal pathway, appear to be connected into a descending motor system: the DLN posterior and lateral to HVC (within a larger region called the caudal lateral nidopallium, NCL) 
projects to LAI directly lateral to RA, which in turn projects to pre-motor neurons (PMN) of the brainstem reticular formation (). Interestingly, the reticular PMN in pigeons, chickens, and ducks also receive a direct projection from the arcopallium; in these species, the reticular PMN laterally adjacent to the nXIIts vocal nucleus projects onto the spinal cord motor neurons that control muscles for wing and leg movements, and when stimulated electrically or with neurotransmitters, induce wing beats, hopping, or walking 
. The parrot and hummingbird posterior vocal nuclei also make a similar descending motor projection () 
In terms of apparent connectivity between posterior and anterior movement-associated areas, the shell of neurons around songbird MMAN and the comparable area in non-songbirds projects to NCL (inclusive of DLN) in a similar manner as MMAN projects to HVC 
; the shell around LMAN projects to LAI in a similar manner as LMAN projects to RA () 
. Differences within the vocal pathway are as follows: unlike HVC's projection to Area X, the adjacent DLN in zebra finches only sends a weak projection to the striatum, whereas the LAI adjacent to RA sends a strong projection to the striatum () and many other areas besides the reticular PMN 
. Likewise, there are more differences in the connectivity between the posterior and anterior vocal pathways of songbirds and parrots than there is within each of the vocal pathways () 
; for example, in both groups, output of the anterior vocal pathway to the posterior vocal pathway is via the MAN-like nucleus, but the input is either from the HVC-like nucleus (songbirds) or the RA-like nucleus (parrots).
Interestingly, in zebra finches, AH sends some of its heaviest cerebral projections to the areas around the anterior vocal pathway nuclei, to the DLN lateral to HVC, and to ventral AI 
; this connectivity pattern is also strikingly similar to the movement associated gene expression we found (b
). This overlap of connectivity and gene expression patterns suggests that AH may transmit somatosensory input into putative anterior and posterior motor pathways adjacent to vocal nuclei, or that the areas adjacent to vocal nuclei are somatosensory instead of motor. If the latter possibility were true, however, there would be no activated cerebral areas left for the motor control of movement. As for the vocal nuclei themselves, they do not require somatosensory input from the syrinx for their vocalizing-driven gene expression and this is one reason why this gene expression has been designated motor-driven 
; the vocal nuclei also show pre-motor neural firing during singing 
. Thus, based on parallels with the vocal nuclei, we hypothesize that the movement-associated areas adjacent to the vocal nuclei will show both pre-motor firing for movement control and somatosensory feedback firing from AH after initiation of movement. These hypotheses on connectivity and activity can be confirmed or falsified in future studies that perform double labeling experiments with injected neural tracers and movement-induced IEG expression, that locally remove cerebral somatosensory input into posterior and anterior movement-associated areas, and that perform electrophysiological recordings during movement. Preliminary electrophysiological studies from our group indicate that the AN area adjacent to zebra finch LMAN has pre-motor neural firing during hopping (Tremere, Pinaud, and Jarvis; Soc. Neurosci. Abstracts, 2007, 221.9).
Although future work still needs to be conducted to decipher motor and/or somatosensory roles of the identified brain areas of this study, our findings help advance future investigations of the avian brain. Since it is difficult to make birds sit absolutely still during a sensory task and obviously impossible during behavioral tasks, general movement-associated activation may distract experimenters' attention from other types of activation. Thus, knowing the brain areas activated during movement will be important for studies using behavioral molecular mapping to identify brain areas involved in specific behaviors, including sensory-motor learning tasks.
Taken together, we speculate that the vocal control circuits, which use auditory information to influence vocal motor output are anatomically adjacent to putative motor control circuits, which use somatosensory information to influence motor output. We suggest that the motor control circuits comprise a general cerebral motor system consisting of two sub-pathways: an anterior and a posterior pathway (b
). Based on parallels with the vocal system, we hypothesize that the adjacent posterior pathway controls the production of movements and the adjacent anterior pathway controls sequencing if not learning of movements. Although, our results do not indicate whether the anterior areas are involved in motor learning (such as learning to walk, fly, or manipulate the beak), as is the case for the anterior vocal nuclei during song learning, like the anterior vocal pathway's activation during singing, the adjacent movement-associated areas are active during production of motor behaviors. Our approach was not specific enough and did not aim to map a possible homunculus organization as seen in the mammalian motor cortex 
, but given that the avian AH and AMD has somatotopic organization 
and that restricted patterns of activation occurred adjacent to the vocal nuclei depending on the types of movements performed, it would not be surprising to find a homunculus-like organization in a putative avian anterior and posterior motor system as well.
A motor theory for vocal learning origin
Based on the above findings and related studies, we propose the following theory: Cerebral systems that control vocal learning in distantly related animals evolved as specializations of a pre-existing motor system inherited from their common ancestor that controls movement, and perhaps motor learning.
Although the results of this study do not prove this theory-the reason for calling it a theory-they support it more than other theories of vocal learning origin; we called our idea a theory as opposed to a hypothesis, because it consists of multiple hypotheses. Others have suggested that forebrain vocal learning systems for learning and production in various species including humans have evolved out of either a pre-existing auditory pathway 
, an auditory-motor system 
, a non-motor cognitive system 
, or de-novo 
. Support for some of these theories in birds are: i) The songbird posterior vocal nuclei are adjacent to and share similar connectivity with the descending auditory pathway 
; ii) The parrot posterior vocal nucleus NLC is surrounded by auditory responsive neurons 
; iii) All vocal nuclei of songbirds show neural firing when hearing song, leading to the motor theory of song perception 
; and iv) The striatum around Area X was thought to show hearing-induced IEG expression 
. Alternative explanations or interpretations can now be offered for these findings. These include that: i) The descending auditory pathway, which is not activated by movement, has connectivity that is similar to movement-associated areas 
(noted here) and thus this may not be a good distinguishing factor; ii) The auditory-evoked neural firing in the songbird vocal nuclei occurs mainly in anesthetized or sleeping birds 
whereas when the birds are awake, firing and IEG induction is mainly motor-driven (singing) independent of auditory input 
; iii) Similar to auditory responses in vocal nuclei of anesthetized songbirds and parrots, the auditory responses adjacent to NLC in parrots 
may reflect sensory input into a motor system, in that parrot SLN around NLC behaves more like DLN adjacent to songbird HVC than to the auditory shelf adjacent to HVC; and iv) The hearing-associated gene activation around anterior songbird vocal nuclei can be explained by animals moving in response to hearing song (this study).
These alternative explanations and interpretations do not mean that auditory or other sensory information does not enter the vocal or putative adjacent motor system. On the contrary, sensory information must enter the systems in order to control sensory-motor guided behavior. For auditory input, the best candidate songbird nuclei so far are the vocal nucleus NIf 
and the auditory CM (MV-L2) area 
in which Av is located. Interestingly, the adjacent PLN and PLMV regions were the only two areas that showed both movement- and auditory-associated activation independent of each other, and thus we speculate that they could possibly represent a pre-existing multimodal brain cluster where auditory information is transmitted to a motor system. Likewise, one view for budgerigars is that the vocal nucleus proposed to be analogous to NIf (LAN 
) receives input from auditory fields L1 and L3 
; but this remains to be further tested and considered with possible dual auditory input from an auditory part of basorostralis in parrots 
Our theory can explain why the cerebral vocal systems are similar across distantly related vocal learning birds (). For many years, there has not been a satisfactory explanation for the finding that songbirds, parrots, and hummingbirds have seven comparable cerebral vocal nuclei that cannot be found in their close vocal non-learning relatives 
. One possible explanation was that supposed vocal non-learners actually have rudimentary vocal learning behavior and rudimentary cerebral vocal nuclei that were then independently amplified in vocal learners (); however, none have been found despite efforts to search for them 
(and this study). Another explanation was that a cerebral vocal pathway existed in a common vocal learning ancestor of vocal learning birds that was then lost multiple independent times in their cousins 
(, green dots). A third and the dominant hypothesis was that each of the three vocal learning bird groups evolved similar cerebral nuclei for vocal learning and production independent of their common ancestor 
(, red dots). Our theory offers a modified view of the independent evolution hypothesis, this being that the three vocal learning bird groups independently evolved similar cerebral vocal systems but that were dependent, i.e. constrained, by a previous genetically determined motor system inherited from their common ancestor (). This pre-existing motor system may be a basic motor system of the avian brain that consists of distinct areas (possibly seven nuclei) distributed into two pathways (posterior and anterior), which in parallel incorporate portions of different cerebral subdivisions (mesopallium, nidopallium, arcopallium, and striatum), each sub-serving a specific function. If true, then such a basic posterior/anterior motor system that controls different non-vocal muscles in parallel pathways via premotor neurons in the brainstem could be used as a template for the evolution of a vocal motor/learning system that controls muscles of the syrinx, taking over control of DM and nXIIts that normally controls innate vocalizations. This hypothesis may be testable with fate mapping and genetic manipulation studies of developing brain circuits.
One potential caveat of our theory is that the posterior vocal nuclei are in different locations in each vocal learning avian order and so are the adjacent movement-associated regions. The biggest relative differences are seen in parrots compared to all the other species (songbirds, hummingbirds, ring doves, chickens, quails, pigeons, and suboscines) we have examined in our studies. One possible explanation is that the posterior motor pathway migrated more anterior-laterally during evolution of the parrot ancestor and that the posterior vocal pathway moved with it or later evolved out of it. Support for this general idea is that the arcopallium in parrots is positioned much further anterior relative to other avian species, although it is still posterior relative to the anterior vocal nuclei. If the motor part of the nidopallium moved with the arcopallium anterior and laterally, then this would suggest that parrot SLN is the homolog of DLN in the other species (). Further, the parrot cerebrum, and the nidopallium in particular, is much larger in brain to body size ratio relative to other species 
. Since the posterior nidopallium also contains sensory integration pathways 
, perhaps such sensory pathways were expanded in parrots displacing the posterior motor pathway anterior and laterally. This idea can be tested by mapping the functional organization of the nidopallium and the connecting arcopallium between the auditory and posterior movement-associated areas in parrots relative to other species. In hummingbirds, the posterior movement-associated areas are in a more similar position relative to songbirds and ring doves, but the posterior vocal nuclei are positioned more lateral instead of medial to the movement areas. Such differences suggest that it is likely that the vocal nuclei in each bird order evolved independently, but from the common ancestor motor pathway substrate ().
We believe that our findings may also have implications for understanding the evolution of brain pathways for vocal learning among distantly related mammals. The phylogenetic distances among vocal learning mammals (humans, bats, sea mammals, and elephants) are similar to those among vocal learning birds 
. Comparative analyses among vocal learning and non-learning mammals 
and between mammals and birds 
, with humans being the only vocal learner for which cerebral vocal (speech) brain regions are known, indicate some analogies between humans and vocal learning birds 
. These include in humans a proposed anterior vocal pathway involving Broca's area, adjacent cortices, the anterior striatum, and anterior thalamus and a posterior vocal pathway that comprises the face motor cortex and its projections to brainstem vocal motor neurons 
. The face motor cortex is within the motor cortex and Broca's area is adjacent to or considered by some to be within the pre-motor cortex 
. An analogous area of the pre-motor cortex in non-human primates, macaques, a vocal non-learner, modulates orofacial, but not laryngeal, movements 
. Further, the mammalian non-vocal motor (posterior) and pre-motor (anterior) pathways follow a connectivity design similar to the songbird and parrot posterior and anterior vocal pathways 
; these are the mammalian descending motor pathway and cortical-basal-ganglia-thalamic motor loops, respectively. Perhaps the evolution of vocal learning brain areas for birds and humans exploited a more universal motor system that predates the split from the common ancestor of birds and mammals, i.e. stem amniotes 
. Such a universal system would be consistent with both proposed hypotheses of avian and mammalian pallial homologies, which are that pallial areas containing the vocal nuclei in birds are homologous to either the mammalian six layered cortex or to the mammalian claustrum-amygdala complex 
, if the latter in mammals were found to consist of a rudimentary motor system. This hypothesis can be strengthened or weakened by studying brain pathways for vocal learning in other vocal learning mammals as well as non-vocal motor pathways of reptiles and amphibians.
At this point, we cannot say in our theory whether the forebrain vocal system formed by using a pre-existing part of a motor pathway as a scaffold or usurped a pre-existing part of the pathway. However, we do not believe that a pre-existing part of a motor pathway was lost. Rather, our theory is in line with previous ideas on evolution of novel brain systems from older systems. For example, Finlay 
suggested that new mammalian cortex areas arise first by an enlargement of an older region and then second by allocating part of that older region to the new function, while the remaining part maintains the old function. This is similar to the idea that new functions can be generated by gene duplications, where a gene is duplicated and one copy is used for a new function while the old copy maintains its function 
. More universally, Ghysen 
argues that vertebrate as well as insect brains have ancient principle sensory and motor circuits with stable functions upon which alterations by gene mutations and embryonic development during evolution are applied to home new functions. These altered circuits may then be uncoupled from the original pathways to allow the novel functions without affecting the original system. Perhaps vocal learning systems have evolved by such a mechanism.
Although our findings led us to propose the above theory, we are not the first to implicate a motor origin for a learned vocal behavior. Based upon a literature summary of studies conducted in humans, Robin Allot in a linguistic conference proceedings 
proposed a “motor theory for language origin” where he argued that language brain areas evolved from a pre-existing motor neural system; however, he did not provide experimental evidence or flesh out the anatomical or mechanistic details of this theory. Lieberman 
proposed that language areas evolved out of a pre-existing cortical-basal-ganglia-thalamic-loop, for which he deemed the basal ganglia part as the reptilian brain. However, we now know that reptilian and avian cerebrums are not made up of only basal ganglia, that vocal learning birds only have part of the vocal system in the basal ganglia, and that spoken language areas may involve more than just this loop 
. Farries 
and Perkel 
proposed in birds and Jarvis 
in birds and humans, that vocal learning pathways in birds and humans may be similar to systems outside of the vocal pathways that intuitively could be motor pathways found in vocal non-learning birds and mammals; but they did not have experimental evidence to corroborate these suggestions. Here we provide evidence that the brain areas adjacent to the vocal systems of all known vocal learning birds function during movement. This poses the question-what makes vocal learning, and spoken language for that matter, special-a question that is often debated 
. We argue that it is a cerebral motor system that controls the vocal apparatus. That is, vocal learners and non-learners have similar auditory pathways, but vocal learners have a unique vocal motor system that gives them the ability to translate auditory signals into vocal signals. Like in birds, it is not clear how the auditory information reaches the vocal motor areas but a dorsal sensorimotor stream from secondary auditory cortex to Broca's area has been one hypothesized system 
Our results are also concordant with the gestural origin of spoken language hypothesis, where the motor learning ability of gestures in humans and non-human primates has been argued to be the precursor behavior for motor learning of speech/language 
. During child development, gesture production appears before speech production and is thought to enhance learning of speech; adults also use limb gestures automatically and often unconsciously during speech production 
. This gesturing hypothesis was one basis for the motor theory of language origin 
. We suggest that, logically, gesturing is controlled by a pre-existing motor system. Gesturing, although not a requirement in our theory, has not been well studied in birds, but many avian species perform other movements such as a courtship dance or wing displays during vocalizing 
. Investigations into the behaviors and neural circuits for movement displays in birds may help shed light onto these ideas. If verified in both birds and mammals, then the evolution of vocal learning brain systems as a specialization of a pre-existing motor system could be a general feature of the vertebrate brain.