Areas In The Avian Telencephalon Important For Vocal Learning

Vocal learning is a rare trait that enables an organism to imitate sounds that it hears. To date, vocal learning is known to exist in only three avian orders, which are thought to have evolved this trait independently (e.g., Brenowitz, 1997; Nottebohm, 1970). There is some evidence to suggest that the phylogenesis of vocal learning arises in parallel with a systematic constraint in the telencephalon (Jarvis et al., 2000). All three orders of vocal learners have vocal learning nuclei in the same relative positions within their respective telencephalons. Furthermore, the connectivity between them is conserved so that there are two distinct pathways: a posterior motor/production pathway and an anterior forebrain pathway (AFP). The posterior motor pathway is directly involved with singing throughout the bird's life. The anterior forebrain pathway is necessary for song learning, but under normal circumstances is not required after the final song is established.

FIGURE 16.1 The connections of the posterior motor pathway and anterior forebrain pathway. The posterior motor pathway (white arrows) begins in the high vocal center and extends to the robust nucleus of the archistriatum via HVC[RA] projection neurons. RA neurons extend to lower motor neurons/respiratory neurons in the songbird throat (see text). For the AFP (black arrows), Area X receives input from the HVC from HVC[X] projection neurons. The spiny striatal cells in Area X extend to pallidal-like cells, also within Area X. The pallidal-like cells leave Area X and input to neurons in the medial nucleus of the dorsolateral thalamus that proceed to send efferents to the lateral magnocellular nucleus of the anterior neostriatum. Bifurcating neurons within LMAN project to RA and back to Area X. The Area X to DLM to LMAN pathway is closed and remains in register (see text).

FIGURE 16.1 The connections of the posterior motor pathway and anterior forebrain pathway. The posterior motor pathway (white arrows) begins in the high vocal center and extends to the robust nucleus of the archistriatum via HVC[RA] projection neurons. RA neurons extend to lower motor neurons/respiratory neurons in the songbird throat (see text). For the AFP (black arrows), Area X receives input from the HVC from HVC[X] projection neurons. The spiny striatal cells in Area X extend to pallidal-like cells, also within Area X. The pallidal-like cells leave Area X and input to neurons in the medial nucleus of the dorsolateral thalamus that proceed to send efferents to the lateral magnocellular nucleus of the anterior neostriatum. Bifurcating neurons within LMAN project to RA and back to Area X. The Area X to DLM to LMAN pathway is closed and remains in register (see text).

The posterior motor pathway encompasses a nucleus in the avian pallium called the high vocal center (HVC) and one of its projections to the nucleus, called robustus archistriatalis (RA). The HVC also contributes to the AFP by sending a projection to Area X, as illustrated in Figure 16.1. Area X is found within the avian equivalent of the striatum. Area X sends an efferent to the dorsolateral anterior thalamic nucleus (DLM) that in turn projects back up to another ventral pallial area called the lateral magnocellular nucleus of the anterior neostriatum (LMAN). The LMAN projection to RA is topographic so that the projections are horizontally layered across RA. This organization is compatible with a myotopic organization of the RA efferents that extend to the throat musculature (Vicario, 1991). LMAN projection neurons bifurcate so that one axon extends to RA, and a collateral extends to Area X. The LMAN projection to Area X forms a feedback loop to give rise to distinct modular domains.

The AFP is a pathway that parts from HVC only to append with the posterior motor pathway at the level of RA. What is more, an additional cell type is present in Area X that resembles those found in the mammalian dorsal pallidum, or globus pallidus (Fames and Perkel, 2002). Therefore, the AFP is recursive and forms a closed cortico-striatal loop: the parts of the loop are Area X, DLM, and LMAN (Luo et al., 2001). Current thinking is that the spiny striatal cell in Area X projects to the pallidal-like cells, which then project to the thalamic DLM, which sends efferents to LMAN. The pallidal-like cells in Area X are far outnumbered by the spiny cells, suggestive of a high degree of convergence. Thus, all of the elements of the mammalian cortico-striatal module are embraced within the avian cortico-striatal counterpart. The loop is topographically organized in a manner that accords with the myotopic architecture of RA.

The avian cortico-striatal-like module may not be precisely analogous to its mammalian counterpart. The mammalian module receives projections from the dorsal pallium, while the avian module includes projections from the ventral and lateral pallium — what will loosely be considered the avian cortex. Nevertheless, from an interval timing perspective, these findings imply that the evolution of integrated sensorimotor behavior may entail the development of a common unit of processing. The modular organization in Area X suggests that these are functional domains that process information via a closed loop in a manner that preserves myotopic organization (Vates and Nottebohm, 1995). The main question is whether this closed topographical loop specialized for the learning and production of birdsong also plays a role in the more general process of interval timing? In order to address how the vocal pathways are involved in sensorimotor learning, we will first begin at the level of motor production. From there, we will ask how the hierarchical structure of a learned song template can be used to shape the song's production.

The zebra finch (Taenopyggia guttata) is the preferred choice of species for those who look at birdsong from a neurobiological perspective. This is partly due to the fact that the zebra finch has one motif (song type) that is syntactically simple and deterministic. The motif is composed of several syllables, each syllable made up of notes. The zebra finch will engage in singing bouts where it will repeat its motif. The variation in the motif over a bout is very small, although this is influenced by different social contexts (Sossinka and Bohner, 1980).

16.4.1 Posterior Motor Pathway

The HVC is thought to be an avian homologue of the mammalian primary motor cortex, layers II and III, and the RA nucleus is thought to be a homologue of layer V, which projects to lower motor neurons (Karten, 1969). Specifically, RA projects to a set of brain stem nuclei that coordinate respiration and vocalization; the most prominent are the tracheosyringeal part of the hypoglossal nerve motor nucleus (nXIIts).

It was established early on that both HVC and RA are required for production of song, as lesions of either HVC or RA result in silent song (Nottebohm et al., 1976). Silent song was distinguished by the observation that the songbird engaged in all other mannerisms that were characteristically associated with singing, but did not produce song itself. These results suggested that both RA and HVC are necessary to produce movement in the songbird's throat musculature. This may not be surprising given that lesions of the mammalian motor isocortex lead to severe deficits in motor performance. It is difficult to find parallels in the mammalian interval timing literature regarding primary motor isocortical lesions. If the behavioral output is severely disrupted, one cannot properly assess deficits in interval timing (see Sasaki et al., 2001).

Single-unit recordings in HVC cells identified as projection neurons to RA (HVC[RA]) exhibit bursting only at a specific time in a motif. That is to say, each HVC[RA] cell that was recorded bursted reliably at exactly one point in the sequence (Hahnloser et al., 2002). In addition, simultaneous recordings in RA revealed that a single RA projection neuron to the throat musculature will burst multiple times during a single motif, but each occasion appears to be tightly time-locked to the single firing of a specific HVC[RA] neuron. Moreover, there is strong evidence supporting a causative relationship between the burst of an HVC[RA] neuron and an ensuing burst in an RA neuron that occurs with a minimal latency. Thus, it would seem that a single HVC[RA] neuron fires dependably at one time in the motif and immediately yields a burst in an RA cell. A single RA cell may burst multiple times during a motif, but presumably each burst is driven by a different HVC[RA] cell that is time-locked to fire at that point in time. It has been proposed that there is a direct temporal map imposed on the various HVC[RA] neurons (Hahnloser et al., 2002). Different subpopulations of HVC[RA] neurons are firing at a predetermined time once per motif, and this drives the firing of different RA cells projecting onto the songbird's throat musculature.

If a syllable is repeated twice within the same motif, will the subpopulation of HVC[RA] cells used to code the syllable's first occurrence be used to induce the same RA cell activity during its second occurrence? In other words, how specific is the activity of HVC[RA] cells during a motif with respect to time? A purely temporal code would imply that a unique subpopulation of neurons would burst fire only once during a given motif's duration. Although this has not been directly addressed, there is sufficient evidence to offer a hypothesis.

Each RA neuron generates a mean of approximately ten bursts for each motif so that 15% of the total RA population is active at any one time during a motif (Hahnloser et al., 2002). Identical patterns generated by a group of RA neurons lead to identical syllables, whereas similar, but not identical syllables that share many notes in common can have drastically different RA patterns. Therefore, it seems likely that if a syllable is repeated twice within the same motif, the same population of HVC[RA] neurons codes the sequence of each syllable (M. Fee, personal communication, June, 2002). The different subpopulations of HVC[RA] neurons that are time-locked for an RA sequence serve as indicators that "point" to a group of RA neurons corresponding to a sound that is intended to be produced at that point in time. This is entirely different from the earlier song code mechanisms that were proposed (e.g., Yu and Margoliash, 1996). Previously, HVC and RA were functionally dichotomized. HVC was thought to code for structure over an interval that corresponded to a syllable, and RA for much shorter durations at the level of a note. In the most recent proposal, the sequence of activity that unfolds over time is inherent to the individual subpopulations of cells in HVC. This sequence of activity is projected onto a "muscle map" in RA to create a song so that HVC and RA are functionally connected; i.e., a temporal code operates on a spatial code. The results presume that the time at which these HVC[RA] neurons are firing in a motif is already ingrained within their respective subpopulations. These conclusions do not focus on how the neurons in HVC[RA] come to be tuned to fire at an appropriate point in time in order to select the sequence of RA neurons corresponding to a particular syllable.

How is the auditory template in the songbird used to guide the formation of connections between the HCV[RA] and RA? Before answering this question, the location of the auditory template in the avian brain must be uncovered.

16.4.2 Is the Auditory Template Contained within the Anterior Forebrain Pathway?

The finding that the AFP is crucial for the development of complex vocalizations within the songbird, but not for maintenance under normal conditions in adulthood, was a significant discovery for those searching for the auditory template in the songbird (Bottjer et al., 1984; Scharff and Nottebohm, 1991). Auditory properties were first described in the HVC of anesthetized zebra finch; however, they were soon found throughout the AFP as well. Additionally, auditory responses have been recorded within RA of the zebra finch. The responses within RA are thought to depend on HVC input due to the fact that lesions of LMAN (also afferent to RA) do not abolish auditory responses (Doupe and Konishi, 1991).

An auditory property is established within a cell if it is observed to preferably respond to one sound stimulus over another. Many neurons in HVC fire in preference to an autogenous song (what will henceforth be referred to as the bird's own song (BOS)) over a conspecific or heterospecific song. With this in mind, the auditory properties in HVC have been considerably elaborated on by exposing the adult bird to different types of sounds such as the BOS. Multiunit recordings of HVC neurons fired vigorously during the BOS, more than what was observed when the bird was presented with its tutor's song or other conspecific song (Margoliash, 1986; Mar-goliash and Konishi, 1985). What's more, often these neurons will exhibit temporal combination sensitivity (Lewicki and Konishi, 1995; Margoliash, 1983; Margoliash and Fortune, 1992). This property illustrates a neuron's sensitivity to two attributes in a complex acoustical stimulus like birdsong: order selectivity and combination sensitivity. Order selectivity is demonstrated when the neuron fires at a higher rate to the BOS than when the BOS syntactic organization is manipulated. This includes playing back the song in reverse (a global song manipulation), reversing the temporal order within the syllables, but maintaining the same syllable sequence (a local song manipulation) or presenting the syllables in reverse order so that the local temporal order within each syllable remains the same (a more specific global manipulation). Combination sensitivity describes an observation that a neuron will fire in response to a combination of certain syllables at a greater rate than what would be predicted by summing the responses to each individual syllable component.

Consequently, the AFPs necessity for song learning, the existence of auditory properties within the AFP that are specific to the juvenile BOS, and association with a posterior pathway that is necessary for motor production called to mind the possibility that the AFP contained the auditory template (e.g., Doupe, 1993; Doupe and Solis, 1997). However, the vast majority of experiments that confirm auditory responses in various song learning nuclei are performed on the anesthetized songbird. It is very difficult to obtain auditory responses from the HVC and other vocal nuclei in the awake bird. The auditory responses recorded in HVC and other vocal nuclei are a product of the behavioral state that is generated by the anesthetic (Capsius and Leppelsack, 1996; Dave et al., 1998; Margoliash, 1997; Schmidt and Konishi, 1998). The auditory properties obtained from HVC[RA] cells in the anesthetized bird compare favorably with recent results displaying the feature of one burst per motif, reminiscent of the "temporal code" (Hahnloser et al., 2002). Thus, the auditory responses may be a reflection of the motor response one would expect in the awake, behaving bird rather than true acoustic-based properties. For that reason, it is equally likely that the AFP represents a motor-like conduit that can use an auditory-related signal generated elsewhere. That is, the AFP may not be the genuine locus of auditory memories, but may serve as a specialized motor pathway that conducts a signal used to facilitate the formation of accurate HVC[RA] and RA synapses in the posterior motor pathway. Indeed, it came as a surprise to some avian researchers that the AFP was active during song production, as revealed by robust gene expression within the AFP nuclei (Jarvis and Nottebohm, 1997).

Another point to consider about the AFP in the context of song development is the importance of auditory feedback. The final song is made to be abnormal after depriving the songbird of auditory feedback by deafening after the memorization phase but preceding the sensorimotor phase (Konishi and Nottebohm, 1969). What is more, the nature of crystallized adult song is not so straightforward. The maintenance of the adult final song requires real-time auditory feedback (Leonardo and Konishi, 1999; Nordeen and Nordeen, 1992; Okanoya and Yamaguchi, 1997). For example, deafening the adult songbird or perturbing the auditory feedback can result in decrystallization of the adult song. On the other hand, lesioning parts of the AFP that disrupt this circuit in a deafened adult prevent song deterioration (Brainard and Doupe, 2000). Therefore, the AFP is also thought to play a role in evaluating auditory feedback.

16.4.3 An Alternative Site for the Auditory Template

We propose that the source of the auditory template resides in areas that are involved in basic auditory processing in songbirds and nonsongbirds. Field L is the primary auditory region in the songbird and is the avian homologue of mammalian auditory cortex. It is a large and laminated structure that is subdivided and composed of multiple nuclei that abut with indistinct borders (Fortune and Margoliash, 1992; Vates et al., 1996). Field L is functionally complicated, abounding with reciprocal connections among its subdivisions, and is organized in a tonotopic manner (Muller and Leppelsack, 1985). One of its subdivisions sends a dense projection to a nearby area known as the caudal nucleus of the neostriatum (NCM). The NCM is thought to be an important component of the song system due to its apparent involvement in song perception. There is a vigorous increase of electrophysiological activity at the onset of the sensory phase in the NCM following conspecific song presentation relative to other songs or simple acoustic stimuli (e.g., tone bursts or white noise) in the awake, behaving zebra finch (Stripling et al., 1997, 2001). More recently, it was found that the presentation of individual syllables to an adult canary gives rise to unique patterns of early gene expression in the NCM (Ribeiro et al., 1998). A

complex stimulus like the syllable appears to be represented by distinct subsets of the NCM cells that are highly integrated and are recruited in a unique manner to encode the stimulus. All of these findings intimate that different syllable-related activity within the NCM may interact in a way that makes the formation of auditory representations possible.

Moreover, subpallial regions appear to be involved with auditory processing. Gene expression studies in both the zebra finch (Mello and Clayton, 1996) and the hummingbird (Jarvis et al., 2000) indicate that the caudaldorsal paleostriatum is active upon hearing birdsong. The position of the caudal paleostriatum (PC) situated under the auditory pallial regions is consistent with it being an auditory area of the striatum. There is evidence that the NCM is reciprocally connected with the PC (Mello and Clayton, 1996), and anterograde tracers indicate innervation from fields L1, L2b, and L3 (Vates et al., 1996). Furthermore, the entire avian striatum, comprising the PC, projects to the dorsal pallidum, and the neurochemical attributes of the projections mimic those striatal efferents in mammals. The pallidal outputs project to the MDNs and thalamus (Medina and Reiner, 1997). Functionally, there is very little else known about the PC; however, its close association with pallial auditory regions and input from the auditory thalamic nucleus (Durand et al., 1992) intimate a fundamental role in audition and may contribute along with the NCM in an interface between song perception and production. Certainly, the connections that include the PC are amenable to a closed cortico-striatal loop, though this has not been confirmed. It should be noted that unlike Area X, the pallidal component is not part of the PC, so the striatal projection is not as local, though the possibility of a pallidal-like cell component still remains.

16.4.4 What Type of Biological Substrate Is Required for Song Learning?

The faithful reproduction of a learned song requires first discriminating the temporal relationships among the sequence of perceived elements in the song. This information should be encoded in a manner that permits the construction of motor sequences that match the auditory template. The comparison between what is produced (realtime auditory feedback) and what ought to be produced (the auditory template) need not be made within the AFP, though the AFP is expected to play a role in transmitting the results of this comparison. One potential source of this comparison signal is in the MDNs, which project to Area X in the avian striatum. The source of this signal is significant because dopamine (DA) is thought to play an important role in song learning, although its specific role is far from understood. During the sensorimotor phase of the zebra finch, DA levels rise intensely and peak throughout the AFP (Harding et al., 1998). Furthermore, DA modulates the excitability of spiny neurons within Area X (Ding and Perkel, 2002).

In mammals, there is substantial evidence to suggest that the MDNs encode both the occurrence and the time of reward as assessed during Pavlovian conditioning. This is indicative of a role in expectation or prediction (Hollerman and Schultz, 1998). During Pavlovian conditioning, the MDNs respond vigorously to the unconditional stimulus (US). However, as learning progresses, there appears to be a transfer of excitatory activity from the US to the onset of the conditional stimulus (CS). This transfer of activity is suggestive of a process where the neuronal activity is coactive with the detection of the earliest stimulus that best predicts a reward. However, if the delivery of reward is delayed following training, there is a depression in activity at the time reward is normally expected, followed by a burst of activity when the reward is delivered. The nature of this inhibitory activity implies that it does "not constitute a simple neuronal response, but reflect[s] an expectation process based on an internal clock tracking the time of predicted reward" (Schultz, 1998, p. 4).

The MDNs are also involved in situations that involve learning of sequential CS. The activity in a population of dopaminergic neurons in response to a US following a series of CS-US pairings progressively decreases and is transferred to the onset of each CS (Schultz et al., 1993). This result occurs as long as each CS maintains the same temporal relationship with the US. In the case of vocal learning, if the comparison signal that is transmitted to Area X in the songbird is a function of a discrepancy in the temporal relationship between the real-time auditory feedback and the auditory representation, it could be used to shape the spiny cell output of Area X. In the mammal, most of the MDN input to the caudate nuclei is neuromod-ulatory, this input shapes the spiny neuron's response to cortical inputs, as opposed to providing direct excitatory or inhibitory inputs. If the avian spiny neurons in Area X of the AFP could be shaped so that they fire at a tempo corresponding to a copy of the auditory template, this process could facilitate the formation of appropriate connections in the posterior motor pathway given the two pathways' connective relationship to one another. However, the major question is whether the physiology of the avian telencephalon is conducive to such a system. At this point, one may question the significance of the cortico-striatal loop motif underlying both the auditory template and the AFP.


The existence of parallel cortico-striatal loops in the songbird that interface through the midbrain dopaminergic system could provide a substrate to oversee the production and transmission of a ratio comparison signal. We suggest that there is a cortico-striatal loop in the auditory center of the songbird, henceforth referred to as the auditory processing module. It is involved with generating a comparison signal through real-time auditory feedback that is transmitted to the AFP (i.e., the second cortico-striatal module) via the MDNs. The major components of the auditory-based cortico-striatal loop are recapitulated in Figure 16.2. In this way, a sequence of stimuli (i.e., syllables) that change in an orderly manner as a function of time can cause the auditory processing module to pass a comparison signal to the AFP, which in turn will influence the appropriate connections made in the posterior motor pathway. We will begin our discussion with the role of the cortico-striatal module in interval timing. This will be followed by a discussion of the auditory processing module, and then the AFP and the outcome on the posterior motor pathway.

0 0

Post a comment