Journal of Neurophysiology

Lesions of an Avian Basal Ganglia Circuit Prevent Context-Dependent Changes to Song Variability

Mimi H. Kao, Michael S. Brainard

Abstract

Trial-by-trial variability is important in feedback-based motor learning. Variation in motor output enables evaluation mechanisms to differentially reinforce patterns of motor activity that produce desired behaviors. Here, we studied neural substrates of variability in the performance of adult birdsong, a complex, learned motor skill used for courtship. Song performance is more variable when male birds sing alone (undirected) than when they sing to females (directed). We test the role of the anterior forebrain pathway (AFP), an avian basal ganglia–forebrain circuit, in this socially driven modulation of song variability. We show that lesions of the lateral magnocellular nucleus of the anterior nidopallium (LMAN), the output nucleus of the AFP, cause a reduction in the moment-by-moment variability in syllable structure during undirected song to the level present during directed song. This elimination of song modulation is immediate and long-lasting. We further show that the degree of syllable variability and its modulation are both attenuated in older birds, in concert with decreased variability of LMAN activity in these birds. In contrast to the requirement of LMAN for social modulation of syllable structure, we find that LMAN is not required for modulation of other features of song, such as the number of introductory elements and motif repetitions and the ordering of syllables or for other motor and motivational aspects of courtship. Our findings suggest that a key function of avian basal ganglia circuitry is to regulate vocal performance and plasticity by specifically modulating moment-by-moment variability in the structure of individual song elements.

INTRODUCTION

A hallmark of well-learned skills is that they exhibit a high degree of trial-by-trial reproducibility. Such stereotyped motor output is often considered an endpoint of motor learning, and residual variability in motor performance may be construed as biological noise that either the nervous system is unable to eliminate or that is below threshold for behavioral importance. In the context of learning and ongoing calibration of motor performance, however, trial-by-trial motor variability plays a potentially critical role; variation in motor output enables motor exploration such that evaluation mechanisms can differentially reinforce patterns of motor activity that produce more desired outcomes and weaken or punish patterns of activity that produce worse outcomes. While such an adaptive function of variability is well established in models of reinforcement-based motor learning (Doya and Sejnowski 2000; Sutton and Barto 1998), whether and how the nervous system actively regulates variability in the context of motor practice and ongoing motor performance is little understood. Here, we studied neural substrates of variability in the motor control of adult birdsong.

Song learning shares many features with other forms of vertebrate motor skill learning (Doupe and Kuhl 1999). Initially, during a sensory learning period, birds memorize the sound of an adult tutor song. Subsequently, during a sensorimotor learning period, young birds begin to vocalize and use auditory feedback to gradually refine their vocalizations until they closely approximate the tutor song. Initial vocalizations are highly variable from one rendition to the next. However, as the bird’s song approaches the end state (a good match to the tutor song), it becomes increasingly stereotyped. In the zebra finch, this process is completed by ∼90 days posthatch (dph), at which point the song is said to be crystallized (Immelmann 1969). Despite the apparent stability of adult zebra finch song, disruptions of feedback, such as deafening, lead to a gradual deterioration of performance, indicating that adult experience continues to contribute to maintenance of song (Leonardo and Konishi 1999; Nordeen and Nordeen 1992). Several studies have suggested that learning and maintenance of song may rely in part on the regulation of song variability for purposes of motor exploration (Doya and Sejnowski 2000; Jarvis et al. 1998; Kao et al. 2005; Ölveczky et al. 2005; Scharff and Nottebohm 1991).

Here, we studied the role of the anterior forebrain pathway (AFP), an avian basal ganglia–forebrain circuit (Perkel 2004), in the regulation of song variability. The AFP is an accessory loop that is interconnected with the motor structures for song production (Fig. 1A) and, hence, is well positioned to influence song. Previous lesion studies have shown that the AFP is critical for song learning and feedback-based vocal plasticity and have suggested a possible role for this pathway in modulating song variability (Bottjer et al. 1984; Brainard and Doupe 2000, 2001; Kao et al. 2005; Ölveczky et al. 2005; Scharff and Nottebohm 1991; Williams and Mehta 1999). In juvenile zebra finches, lesions or inactivation of the lateral magnocellular nucleus of the anterior nidopallium (LMAN), the output nucleus of the AFP, disrupt song development and induce premature stereotypy, resulting in highly repetitive, simplified songs (Bottjer et al. 1984; Ölveczky et al. 2005; Scharff and Nottebohm 1991). In adult zebra finches, lesions of LMAN have little overt effect on previously learned song, but they prevent song plasticity induced by experimental manipulation of auditory and/or proprioceptive feedback (Bottjer et al. 1984; Brainard and Doupe 2000, 2001; Nordeen and Nordeen 1993; Scharff and Nottebohm 1991; Williams and Mehta 1999). Consistent with a role in regulating song plasticity, LMAN neurons exhibit robust singing-related activity that is correlated with song structure (Hessler and Doupe 1999a,b). Moreover, manipulation of neural activity in LMAN during singing can alter the spectral structure of individual song elements, or syllables, on a moment-by-moment basis (Kao et al. 2005). These findings suggest that a key function of the AFP in regulating vocal plasticity may be to modulate ongoing patterns of activity and synaptic connections in the motor pathway (Jarvis et al. 1998; Kao et al. 2005; Kittelberger and Mooney 1999; Scharff and Nottebohm 1991). Such a modulatory influence could serve to introduce variability in the motor pathway and subsequent song, which is critical in reinforcement learning (Brainard 2004; Doya and Sejnowski 2000; Kao et al. 2005; Ölveczky et al. 2005; Sutton and Barto 1998).

FIG. 1.

Schematic model for the contribution of an avian basal ganglia–forebrain circuit to social influences on song production. A: song system nuclei. Motor pathway (gray) is required for normal song production throughout life and includes HVC (used as a proper name; Reiner et al. 2004), robust nucleus of the arcopallium (RA), and tracheosyringeal portion of the hypoglossal nucleus (nXIIts). Anterior forebrain pathway (AFP; black) is necessary for vocal motor plasticity but is not required for song production. AFP includes Area X, which is homologous to mammalian basal ganglia, medial nucleus of the dorsolateral thalamus (DLM), and lateral magnocellular nucleus of the anterior nidopallium (LMAN). B: when a male sings alone (undirected song), variability in patterns of LMAN activity may introduce variability into activity in the premotor nucleus RA, resulting in variable song output. In contrast, when a male sings to a female (directed song), stereotyped patterns of LMAN activity may facilitate reproducibility in activity of RA neurons and subsequent song output. C: according to this hypothesis, lesions of LMAN should eliminate socially driven differences in song variability.

To examine the hypothesis that LMAN actively introduces variability into patterns of activity in the motor pathway, we focused on naturally occurring, context-dependent regulation of variability in adult song structure. Song serves as a courtship signal in zebra finches, and previous studies have shown that a variety of aspects of song structure, including its variability, are sensitive to social context (Kao et al. 2005; Sossinka and Böhner 1980). When a male zebra finch sings alone (undirected song), variability in syllable structure is consistently greater than when he sings to a female (directed song) (Kao et al. 2005; Ölveczky et al. 2005). Moreover, these context-dependent changes in song structure are associated with context-dependent changes in neural activity in LMAN. Singing-related activity in LMAN is greater in magnitude and more variable in pattern across repeated renditions during undirected song than during directed song (Hessler and Doupe 1999b; Jarvis et al. 1998; Kao et al. 2005). These findings suggest that greater variability in LMAN activity during undirected singing may actively generate variability in ongoing activity in the song motor pathway and subsequent song output (Fig. 1B). Alternatively, or in addition, stereotyped patterns of LMAN activity during directed song may facilitate reproducibility in motor pathway activity and song.

To examine how LMAN modulates song variability, we compared songs produced in the two social contexts before and after lesions. We previously reported that lesions of LMAN eliminate context-dependent differences in the variability of syllable structure (Kao et al. 2005). Here, we further study the nature of the effects of LMAN lesions. If LMAN actively generates variability in undirected song, lesions should reduce or eliminate variability in undirected song (Fig. 1, B and C). Alternatively, if LMAN actively reduces variability during directed song, lesions should cause an increase in variability in directed song toward the level present in undirected song. In addition, to test the specificity and nature of the role of LMAN in modulating variability, we measured the influence of LMAN lesions on other features of song and behavior that are known to differ with social context. In particular, we asked whether lesions specifically eliminate the social modulation of individual syllables or whether they more generally eliminate modulation of the entire suite of behaviors that differentiate directed and undirected song (Sossinka and Böhner 1980). Finally, to further test the relationship between LMAN activity and song variability, we measured these variables in older adult birds, whose songs are less plastic than those of young adults (Brainard and Doupe 2001; Lombardino and Nottebohm 2000). Our findings support a role for LMAN in actively introducing variability to specific aspects of song structure during undirected song.

METHODS

Subjects

Thirty male zebra finches (Taeniopygia guttata) were used for experiments. All birds were raised in individual breeding cages with their parents and siblings until ≥60 dph. Birds were selected on the basis of their size, singing frequency, and song complexity, and were housed individually in sound-attenuating chambers (Acoustic Systems, Austin, TX). All procedures were performed in accordance with protocols approved by the UCSF Institutional Animal Care and Use Committee.

Surgical procedures

Before all surgical procedures, birds were deprived of food and water for 1 h and anesthetized with an intramuscular injection of Equithesin (0.85 g chloral hydrate (Sigma, St. Louis, MO), 0.21 g pentobarbital (Abbott Laboratories, North Chicago, IL), 0.42 g MgSO4, 2.2 ml 100% ethanol, and 8.6 ml propylene glycol to a total volume of 20 ml with water). After surgery, all skin incisions were sealed with a cyanoacrylate adhesive.

LESIONS.

Bilateral electrolytic lesions were stereotaxically targeted at LMAN, with five penetrations per side and one or two current injections per penetration (50 or 100 μA for 60 s). The amount of LMAN that was removed bilaterally ranged from 40 to 100% (median lesion size: 80%; see Table 1; n = 5 birds), as confirmed by histology at the end of experiments. “Sham” lesions were entirely dorsal to LMAN (Table 1; n = 2 birds). In all birds, the medial magnocellular nucleus of the anterior nidopallium (MMAN) remained intact. Lesions were made in birds between 101 and 123 dph (n = 7 birds). We did not find any correlation between the extent of the lesions and their effects on song output using a variety of measures, including the average degree of modulation of fundamental frequency by social context after lesions of LMAN (P = 0.69) and the average degree of modulation of song tempo by social context after lesions of LMAN (P = 0.49). Moreover, the reported effects of LMAN lesions on the song of the bird with the smallest lesions ranked third overall (median) for changes in the variability of fundamental frequency (70% lesion on the left; 10% lesion on the right as well as a visible reduction in CGRP labeling in the lateral RA on the left and dorsal RA on the right). Hence, we pooled all lesion birds in the subsequent analyses. Restriction of the data set to the four birds with the largest lesions (mean lesion size of 88%) did not change the significance of any of the reported results.

View this table:
TABLE 1.

Extent of lesions

CHRONIC IMPLANTS.

Electrodes were implanted chronically as described previously (Hessler and Doupe 1999a). Briefly, a lightweight microdrive (UCSF and Caltech Machine Shops) carrying two tungsten electrodes (2–5 MOhm) insulated with epoxylite (FHC, Bowdoinham, ME) or glass (A. Ainsworth, Northhamptonshire, UK; Merrill and Ainsworth 1972) was positioned stereotaxically such that the electrode tips were ∼700 μm above LMAN. A reference ground electrode (uninsulated tungsten electrode, A-M Systems, Carlsborg, WA) was implanted such that its tip was located within ∼2 mm of the targeted LMAN. The microdrive and connector socket (FHC) were secured to the skull with epoxy (Devcon, Wood Dale, IL) and dental cement (Dentsply, Milford, DE), and a protective cap was fixed around the microdrive. All electrodes were implanted in the right hemisphere.

Anatomy

After the final recordings, birds were deeply anesthetized with Metofane (Schering-Plough, Union, NJ) and transcardially perfused with 0.9% saline, followed by 3.7% formaldehyde in 0.025 M phosphate buffer. Brains were postfixed for 4 h, cryoprotected, and cut coronally in 40-μm sections with a freezing microtome. Every third section was stained with cresyl violet acetate, and a second set of intervening sections was reacted with an antibody to calcitonin gene-related peptide (Fig. 2B ; CGRP; Sigma, St. Louis, MO) (Bottjer et al. 1997; Brainard and Doupe 2001).

FIG. 2.

Experimental design and song system nuclei in intact and lesioned birds. A: interleaved bouts of directed and undirected song were recorded weekly before and after lesions at each time-point indicated by tick marks. B: coronal sections through AFP (top) and arcopallium (bottom) labeled with an antibody against calcitonin gene–related protein (CGRP; Bottjer et al. 1997). Electrolytic lesions were stereotaxically targeted either at LMAN or at a site dorsal to LMAN (sham). In a sham control (left), LMAN, its axonal projections to target nuclei RA and Area X, and medial MAN (MMAN) are all immunolabeled. In a lesioned bird (right), CGRP labeling in LMAN is absent, and there is a corresponding loss of labeled axonal projections to RA and Area X. MMAN remains intact.

Sound recording

Acoustic signals were recorded by a microphone located above the birdcage and filtered between 200 and 9,000 Hz (Krohn-Hite, Avon, MA). Custom-written acquisition software (C. Malek and A. Leonardo, Caltech, and C. Roddey, UCSF) recorded the acoustic signals before and after the sound amplitude crossed a threshold level. The threshold level was determined experimentally to ensure detection of all vocalizations (calls and songs) and to minimize detection of other sounds, such as the bird flapping its wings. Each bird’s behavior was monitored and recorded by a video camera.

For all experiments, undirected song was recorded when the male was isolated in a sound-attenuating chamber. The amount of undirected song was often greater when the experimental bird could hear the calls of other birds outside the recording chamber. To elicit directed song, one or more female zebra finches was presented in a separate cage to the male zebra finch being recorded. The recorded bird usually moved to the edge of its cage and sang while facing the female(s). Each female presentation lasted for ≤2 min, regardless of whether or not the male sang, and songs were classified as directed only when the male bird faced the female(s). Bouts of directed song were interleaved with undirected song during each recording. Recordings lasted between 30 min and several hours depending on the amount of singing.

For experiments examining the effects of lesions of LMAN on context-dependent differences in song, interleaved bouts of directed and undirected song were recorded as shown in Fig. 2A. Recordings were made in each bird once a week for 2–3 wk preceding surgery, again on the day before bilateral lesions, and after lesions as follows: birds were recorded every day after lesions until enough songs were collected in both social conditions to characterize initial effects of lesions (1–3 days) and once a week for 4 wk after the lesions. A final recording was made during the eighth week after lesions (n = 5 birds with LMAN lesions; n = 2 birds with sham lesions). Three additional age-matched control birds were also recorded weekly. The range of ages of the birds at the first recording was 88–102 dph. Data from a subset of these recordings have been presented in a brief report (Kao et al. 2005).

To study the effect of age on context-dependent differences in song structure, we recorded songs from an additional 6 birds >4 yr old and 11 birds <6 mo old.

Physiological recording

In a separate group of birds (n = 7), we recorded neural activity at multiple sites in LMAN (1–6 sites/bird) in directed and undirected conditions. During each recording, a flexible lead terminating in a small operational amplifier circuit was connected to the socket on the bird’s head, and the other end was connected to a commutator (Caltech Machine Shop). Electrodes were positioned at sites in LMAN where action potentials of single and multiple neurons could clearly be differentiated from the background neural activity (spike amplitudes ranged from 300 μV to >1 mV, peak-to-peak; Hessler and Doupe 1999a). The neural activity signal passed through the commutator to a differential amplifier (A-M Systems) and was filtered between 300 Hz and 10 kHz. The acoustic signal was recorded as described above. Custom-written software (A. Leonardo, Caltech, and C. Roddey, UCSF) recorded the acoustic and neural signals, and the bird’s behavior was monitored and recorded by a video camera.

Simultaneous song and neural recordings were made at intervals of 1d to several weeks over a period of weeks to months. Neural activity was recorded during nonsinging and singing periods in both undirected and directed conditions. After completion of recordings in each bird, small electrolytic lesions (30 μA for 10 s) were made at previously recorded sites. Locations of recording sites were confirmed in 40-μm Nissl-stained brain sections by their positions relative to the depth of the marker lesions.

Behavioral analysis

SONG STRUCTURE.

Zebra finch song can be classified into three levels of organization: syllables, which are individual song elements separated by silent intervals ≥5 ms in duration; motifs, which are stereotyped sequences of syllables; and bouts, which are defined as periods of singing separated by silent intervals ≥2 s in duration (Sossinka and Böhner 1980). Song bouts usually consist of a series of introductory elements followed by a variable number of repetitions of the same motif. Song bouts may either be aimed at another bird (directed) or sung when the male is alone or not orienting toward any other bird in particular (undirected) (Dunn and Zann 1996; Sossinka and Böhner 1980).

ANALYSIS OF SYLLABLE STRUCTURE.

To characterize differences in the structure of individual syllables, we measured the fundamental frequency (FF) of syllables that have constant frequency components (Kao et al. 2005). For a particular syllable, we calculated the autocorrelation of a segment of the sound waveform that had clear harmonic structure (median segment used to calculate FF: ∼50% of the total syllable duration; range = 30–85%). The FF was defined as the distance, in frequency, between the zero-offset peak and the highest peak in the autocorrelation function (ACF) within a range of time lags corresponding to fundamental frequencies typical of zebra finch syllables. To improve the resolution of the frequency estimates, we performed a parabolic interpolation of the peak of the ACF (de Cheveigné and Kawahara 2002). This algorithm was applied to syllables that had clear harmonic structure with a well-defined FF. To examine differences between directed and undirected songs, the FF was calculated for a minimum of 16 renditions (range: 16–90 renditions) in each behavioral context.

ANALYSIS OF OTHER SONG FEATURES.

We analyzed four other features of song that previously have been reported to differ across social context (Sossinka and Böhner 1980): number of introductory elements, number of motifs per bout, the stereotypy of syllable ordering, and song tempo.

To count the number of introductory elements per song bout, we started with the introductory element preceding the first syllable in the song motif and counted backward until there was ≥500 ms of silence or a different type of vocalization, such as a loud distance call. Using this method, we found consistent differences in the number of introductory elements between directed and undirected song bouts (see results). For the number of motifs per bout, we counted all motifs in which at least the first half of the motif was sung because zebra finches often truncate the last motif in a song bout (e.g., if the canonical motif was abcdef, all motifs that included at least abc were counted).

The stereotypy of syllable ordering was quantified by using measures of sequence linearity and sequence consistency, similar to those originally described by Scharff and Nottebohm (1991). We followed the practice of recent studies that have examined stereotypy of syllable sequencing and excluded introductory elements from these measures (Foster and Bottjer 2001; Ölveczky et al. 2005; Zevin et al. 2004).

Sequence linearity quantifies the different possible transitions that can be observed after each unique syllable of song. In this study, we used the internal linearity measure of Foster and Bottjer (2001), which excludes variability in the syllable that precedes song terminations. Internal linearity is defined as Math For example, if a bird always sings the motif abcde, there are five unique syllables and four unique transitions, and the internal linearity yields a value of 1.00. Any variation in the sequencing of syllables within the motif will yield a value <1.00. However, songs in which motifs are terminated after more than one syllable still receive a score of 1.00 (i.e., occurrences of truncated motifs abc and abcd would not reduce the internal linearity score). Because internal linearity excludes variability in song terminations, it always yields a value that is equal to or higher than the sequence linearity score calculated by Scharff and Nottebohm (1991).

To explicitly characterize any context-dependent differences in premature song terminations, we calculated the percentage of syllable types within a motif that were followed by song terminations. For example, if a bird’s typical motif was abcde, but the truncated versions abc and abcd were also observed, the percent of syllables that terminate song would be 60 (3/5).

To measure the frequency with which the main motif sequence occurred, we used the sequence consistency score of Scharff and Nottebohm (1991), excluding variability in introductory elements. Sequence consistency measures the proportion of syllable transitions that conform with the most frequent, or dominant, transition for a given syllable and is defined as Math Complete consistency yields a score of 1.00. Because transitions to the end of a song bout are included in this measure, it is influenced by variability at the end of songs.

Because these measures of sequence stereotypy are sensitive to the number of songs that are analyzed, we calculated linearity, consistency, and percent syllables terminating song from randomly selected subsets of recorded songs that were matched to have equal numbers of motif initiations in the directed and undirected conditions (mean and range of motifs per bird: 67; 36–125).

To characterize any context-dependent differences in song tempo, or the rate at which song is delivered, we measured for each bird the durations of complete motifs during directed and undirected songs. For each recording, we computed the mean motif duration for both behavioral conditions.

ANALYSIS OF COURTSHIP BEHAVIOR.

Directed song serves as a courtship signal and is often accompanied by a rhythmic, pivoting dance that includes orienting toward and approaching the female, hopping to and fro, head-tail twists and bows, beak wipes, and a characteristic posture (reviewed in Zann 1996). To characterize these motor aspects of courtship, we scored video clips of directed and undirected singing. Video clips of directed and undirected song bouts (∼30–60 s) recorded 1 day prelesions and 7 days postlesions were interleaved, and the clips were organized in blocks by bird because the exact form of the courtship dance varies across males. Two observers highly familiar with zebra finch behavior and blind to the experimental manipulation of each bird (lesions of LMAN or sham lesions) judged the degree of arousal and the vigor of the courtship display of each male using a scale of 0 (relatively inactive and no dance movements) to 3 (highly aroused courtship display). The observers were instructed to note the following features of the courtship dance: the male’s posture, orienting toward the female, approaching the female, hopping to and fro, and beak wiping. During undirected singing, when no female was present, scores could be >0 if the male sang in the same position or oriented in the same direction as it did during directed singing. Scores were averaged across song bouts in a particular social context on each day to derive a mean behavioral score. Scores were averaged across the observers.

Analysis of neural signals

Analysis of singing-related activity in LMAN was performed as described previously (Hessler and Doupe 1999a,b). Briefly, rectified, smoothed neural activity waveforms were aligned using a template for the amplitude envelope of each bird’s motif. Both the mean activity level and the coefficient of variation of activity across motif renditions were calculated.

RESULTS

Previous studies have characterized several differences in song depending on the social context in which it is produced (Kao et al. 2005; Ölveczky et al. 2005; Sossinka and Böhner 1980). Sossinka and Böhner (1980) reported that when a male sings to a female (directed song), song is preceded by more introductory elements per bout, includes more repetitions of a stereotyped sequence of syllables, or motif, per bout, is more stereotyped in the sequencing of syllables, and is delivered at a faster tempo than when the male sings alone or to no particular audience (undirected song; Figs. 3A and 8A). In addition, stereotypy in the structure of individual song elements is greater during directed song than during undirected song (Fig. 3B; Kao et al. 2005). Finally, directed song is often accompanied by changes in the orientation, posture, and position of the male, which collectively constitute a courtship dance (Williams 2001; Zann 1996). In the following sections, we describe the contribution of LMAN to each of these context-dependent changes to song and behavior.

FIG. 3.

Context-dependent differences in song output. A: spectrogram (plot of frequency vs. time) of 1 bout of song recorded when a male was alone [undirected song (US); top] and when it sang to a female [directed song (DS); bottom]. Song bouts consist of a series of introductory elements followed by a variable number of repetitions of a stereotyped sequence of syllables, or motif (abc, indicated by bars). Number of introductory elements and motifs per bout are typically lower during US than during DS (Sossinka and Böhner 1980). B: variability in syllable structure is greater in the undirected condition. Spectrogram of the same bird’s motif (abc; top). Syllables a and c contain constant frequency components with a well-defined fundamental frequency (FF). Dotted lines indicate portion of syllable for which FF was measured. Histograms of the FF for indicated portions of syllables a (left) and c (right). Variability in the FF was greater during US than during DS (syllable a: SD: 8.8 vs. 3.4; P < 0.0001; syllable c: SD: 7.2 vs. 2.5; P < 0.0001, F test for equality of variance).

Context-dependent differences in variability of syllable structure require LMAN

To assess the contribution of LMAN to ongoing motor performance and song plasticity, we first characterized the spectral structure of individual syllables during directed and undirected song. We focused our analysis on the fundamental frequency of syllables that contain constant frequency components with clear harmonic structure. FF is a learned parameter of song that is precisely controlled (Tchernichovski et al. 2001) and robust to differences across recording conditions. In addition, microstimulation in LMAN during singing can induce changes in FF on a moment-by-moment basis, supporting the idea that neural activity in LMAN modulates ongoing performance of this vocal feature (Kao et al. 2005).

As previously reported (Kao et al. 2005), we found that in normal adult zebra finches, there was a striking modulation in the variability of FF by social context. Figure 3B shows the variability of FF for two syllables (a and c) produced by an adult male in interleaved bouts of directed and undirected song. As shown here, the distribution of FF for an individual syllable during undirected song was typically broader than the distribution of FF for the same syllable produced during directed song.

In control and sham-lesioned birds, context-dependent differences in the variability of syllable structure were extremely robust. Each symbol in Fig. 4A shows the variability in FF expressed as the coefficient of variation (CV) for a single syllable produced in interleaved recordings of directed and undirected song. Data are shown for all measured syllables in all recording sessions from two control and two sham-lesioned birds. The diagonal line indicates equal variability under the two social conditions. For all groups of birds, the data points are distributed significantly above the diagonal, indicating greater variability during the undirected condition (control: 2–6 syllables/bird; 9 recording sessions; n = 72; presham-lesion: 2–4 syllables/bird; 3 recording sessions; n = 18; and postsham-lesion: 2–4 syllables/bird; 6 recording sessions; n = 34; P < 0.0001, paired sign test).

FIG. 4.

Lesions of LMAN reduce the moment-by-moment variability in FF during undirected song to the level present during directed song. Variability in FF during DS vs. variability in FF during US. Each symbol represents CV of FF for 1 syllable in both social contexts during 1 recording session. A: context-dependent differences in FF were present in control birds (open black circles; n = 9 syllables in 2 birds) and in birds before (open red triangles) and after sham lesions (filled red triangles; n = 6 syllables in 2 birds). B: data are plotted for 19 syllables for all recordings before (open symbols) and after (filled symbols) lesions of LMAN. Black Xs denote means for populations. Note that after lesions of LMAN, the absolute level of variability in FF during undirected song was reduced to the level present during directed song (downward shift in mean after lesions of LMAN).

Lesions of LMAN completely eliminated context-dependent differences in the variability of FF that were present in normal adults. Figure 4B shows variability in FF before and after bilateral lesions of LMAN for all measured syllables in all recording sessions from five birds. Before lesions (open symbols), there was a robust modulation of syllable variability by social context; data points lie above the diagonal line (2–6 syllables/bird; 3 recording sessions; n = 57; P < 0.0001, paired sign test), indicating greater variability during undirected song. After lesions (filled symbols, 2–6 syllables/bird; 6 recording sessions; n = 111), the data points were distributed along the diagonal, and there was no longer a significant modulation of variability by social context.

The elimination of context-dependent modulation of syllable variability by lesions of LMAN was rapid and long-lasting. Figure 5 shows the variability of FF at each experimental time-point for both undirected (open bars) and directed conditions (shaded bars). For control and sham-lesioned birds, there were significant context-dependent differences in song variability for all experimental time points (Fig. 5A). In contrast, context-dependent differences in the variability of syllable structure were absent in the first recordings postlesions (1–3 days after) and remained absent for as long as birds were followed (≤8 wk; Fig. 5B).

FIG. 5.

Lesions of LMAN eliminate context-dependent differences in variability in FF. Bars indicate mean CV of the FF (±SE) for syllables during US (open bars) and DS (shaded bars). Data are plotted for the same syllables in the same birds across all recording sessions. A: summary for all syllables in birds with sham lesions and in age-matched control birds that were recorded weekly. Context-dependent differences in song variability persisted in control birds and in birds with sham lesions (n = 15 syllables in 4 birds; P < 0.01 for each recording, paired sign test). B: summary for all syllables before and after lesions of LMAN. Bilateral lesions of LMAN abolish the socially driven differences in song variability by reducing variability in undirected condition to the level present in the directed condition (19 syllables in 5 birds; P < 0.0001 for each recording prelesions, paired sign test; no significant differences in song variability were found for any recording postlesions). Variability in FF during US was significantly lower in the 1st recording postlesion compared with that 1 day before lesions (P < 0.005, paired sign test). For this and all subsequent figures, ***P < 0.0001, **P < 0.005, and *P ≤ 0.01.

The loss of context-dependent differences in variability could be attributed primarily to a reduction in the variability of syllables produced during undirected song to the level present during directed song. This is shown in Fig. 4B by the predominantly downward shift in the population of syllables to the diagonal line (open and filled X indicates population means for pre- and postlesion recordings, respectively). Indeed, lesions of LMAN caused a significant decrease in the variability of undirected songs (Fig. 5B, P < 0.005, paired sign test for 1 day prelesions vs. 1st recording postlesions) but had no effect on the variability of directed songs (Fig. 5B; P > 0.05, paired sign test). Moreover, the residual variability present in the undirected songs in the first recording postlesions was not significantly different from the variability present in directed songs 1 day prelesions (Fig. 5B, P > 0.05, paired sign test). These data are thus consistent with a model in which greater variability in syllable structure during undirected song is actively introduced by LMAN (Fig. 1, B and C).

Lesions of LMAN, however, did not abolish all variability in syllable structure. We did not observe a decrease in the minimum level of variability in FF (present during directed song) after lesions (Figs. 4B and 5B). This suggests that variability in FF during directed singing, when the patterns of activity in LMAN are stereotyped across repeated renditions of the same motif, represents a residual level of variability in syllable structure in adult birds. This residual variability may reflect motor constraints in the periphery, such as limits in the efficacy of the vocal musculature. Alternatively, other sources of variability may exist elsewhere in the brain, such as variability intrinsic to the motor structures for song production.

Courtship behavior is not affected by lesions of LMAN

The effect of lesions of LMAN on song variability raises the question of whether or not other behaviors of males are grossly affected by lesions. Are the context-dependent differences in syllable variability eliminated by lesions of LMAN simply because the males no longer court females? To examine this issue directly, we scored video clips of directed and undirected bouts of song recorded 1 day prelesions and ∼7 days postlesions. In contrast to undirected song, directed song is often accompanied by a rhythmic dance that includes approaching the female, pivoting the body from side to side, and changing the head position (e.g., bowing and beak wiping), partly to display to the female the physical traits that are specifically male (Williams 2001; Zann 1996).

To assess the courtship behavior of males, human observers who were familiar with zebra finch behavior but blind to the experimental manipulation were asked to judge the degree of arousal and the vigor of the courtship display of each male using a scale of 0 (relatively inactive and no dance movements) to 3 (highly aroused courtship display). The scores were averaged across song bouts in each social context before and after lesions to assess the degree to which each male’s courtship behavior had changed. Figure 6 shows the difference in the behavior of male zebra finches between social contexts. One day before lesions of LMAN, all birds exhibited robust courtship displays when a female was present but were less active when they were alone (Fig. 6, left; n = 5 birds prelesions of LMAN and 2 birds presham lesions). One week after lesions of LMAN, the males continued to vigorously court females (Fig. 6, right), and their behavior was similar to that of birds with sham lesions. Thus the elimination of context-dependent differences in song variability by lesions of LMAN cannot be attributed simply to a reduction in the motivation of males to court females. Moreover, these results suggest that the modulation of the courtship display by social context is mediated by structures other than the AFP.

FIG. 6.

Lesions of LMAN do not abolish other courtship behaviors. Summary for all birds of the average level of arousal and/or vigor of courtship display in each social context 1 day before and 1 wk after lesions of LMAN (○) or sham lesions (▪). Males continued to vigorously court females after lesions of LMAN and sham lesions. Lines connect data points for each bird across social context.

Other song features modulated by social context and their dependence on LMAN

This study (Figs. 4 and 5) and other recent work (Kao et al. 2005; Ölveczky et al. 2005) have shown a robust modulation of the variability of syllable structure by social context. In addition, a previous study by Sossinka and Böhner (1980) examined whether other features of song structure are modulated by social context. They reported context-dependent differences in four parameters of song: 1) the number of introductory elements per bout was consistently greater in directed versus undirected song (7/7 birds); 2) the number of motifs per bout was consistently greater in directed versus undirected songs (6/7 birds); 3) the regularity of the sequence of syllables within a motif appeared to be greater during directed songs than during undirected songs (2/3 birds); and 4) directed songs were usually produced at a faster tempo than undirected songs (6/8 birds). It has been suggested that such differences in song may derive from the differential activation of neurons in the AFP across social context (Jarvis et al. 1998). We therefore examined each of these song features to determine the degree to which it is modulated by social context and whether this modulation is affected by lesions of LMAN.

NUMBER OF INTRODUCTORY ELEMENTS.

Consistent with Sossinka and Böhner (1980), we found a robust modulation of introductory elements by social context. Figure 7B shows the social modulation of the number of introductory elements in interleaved bouts of directed and undirected song produced by one adult male before lesions of LMAN. In this case, the average number of introductory elements preceding song was approximately twice as large during directed versus undirected song. In all birds, before lesions of LMAN, the number of introductory elements was significantly greater during directed song than during undirected song (Fig. 7D, left; P < 0.0005, Mann-Whitney U test for comparisons within 1 recording session for each bird).

FIG. 7.

Context-dependent differences in higher-level features of song are not affected by lesions of LMAN. A: schematic of US and DS bouts. B and C: frequency distributions of number of introductory elements per motif and number of motifs per bout during US (open bars) and DS (shaded bars) for 1 recording session pre-LMAN lesion (bird pur24w55). Both the number of introductory elements and the number of motifs per bout were significantly lower during US than during DS (P < 0.0001, Mann–Whitney U test). D and E: number of introductory elements and number of motifs per bout remain higher during DS after lesions of LMAN. Each point represents the average number of introductory elements per song bout or average number of motifs per bout for 1 recording session, and lines connect data points for the same bird across behavioral contexts (n = 5 birds). Bars indicate group means; error bars indicate SE. Data plotted in B and C are indicated by closed squares. F and G: context-dependent differences in higher-level features of song are also present in older adult birds (same convention as above; n = 6 birds).

Lesions of LMAN did not affect the modulation of introductory elements by social context. For each of the birds whose song was characterized before lesions, we counted introductory elements in songs produced 1 wk after lesions of LMAN (Fig. 7D, right). We found that the number of introductory elements continued to be significantly greater during directed songs than during undirected songs for all birds (P < 0.0005; Mann-Whitney U test), indicating that signals from LMAN are not required for social modulation of this song feature.

NUMBER OF MOTIFS.

Figure 7C shows the social modulation of the number of motifs in interleaved bouts of directed and undirected song produced by one adult male before lesions of LMAN, and Fig. 7E summarizes data from each of the birds. Our results were again consistent with those of Sossinka and Böhner (1980). We found a consistent increase in the number of motifs per bout in directed song versus undirected song. This difference was present in all birds and achieved significance for four of five birds (Fig. 7E, left; P < 0.005, Mann-Whitney U test for comparisons within 1 recording session for each bird).

Lesions of LMAN did not affect the social modulation of numbers of motifs per bout. One week after lesions, directed songs continued to have more motifs per bout than undirected songs in four of five birds (Fig. 7E, right), indicating that signals from LMAN also are not required for social modulation of this higher-level song feature.

STEREOTYPY OF SYLLABLE SEQUENCING.

Sossinka and Böhner (1980) also reported a tendency for the sequence of syllables within a motif to be more stereotyped during directed song than during undirected song. This phenomenon was less robust than the other socially driven changes to song (observed in 2/3 birds), and the study did not provide any quantitative analysis of the variability of syllable sequence. From the example shown in that study (Fig. 5 of Sossinka and Böhner 1980) and the corresponding description, it is apparent that the principle difference observed in the stereotypy of syllable sequencing was greater variability in the transitions that could occur after a given syllable during undirected song than for the same syllable during directed song.

To quantify the stereotypy of syllable sequencing, we used measures of sequence linearity and sequence consistency, which have been used successfully for this purpose in previous studies of zebra finch song (Foster and Bottjer 2001; Ölveczky et al. 2005; Scharff and Nottebohm 1991; Zevin et al. 2004). Sequence linearity quantifies the number of types of transitions that can follow a given syllable but is insensitive to how frequently these different possible transitions occur. For this measure, we followed recent studies and included only transitions between two syllables of the bird’s motif and did not include transitions from syllables to song terminations (Foster and Bottjer 2001; Ölveczky et al. 2005; Zevin et al. 2004). Sequence consistency quantifies how often the observed transitions from a given syllable conform to the most common transition for that syllable. Both of these measures should be sensitive to the types of changes in syllable sequencing reported by Sossinka and Böhner (1980). Indeed, for the example bird shown in that study, this measure of sequence linearity increases from 0.64 during undirected song to 0.82 during directed song.

We calculated these two measures for interleaved directed and undirected songs from control birds and from experimental birds before lesions (Table 2, n = 10). In contrast to the report of Sossinka and Böhner (1980), we found that, for both measures, there was no difference in the stereotypy of syllable sequencing between directed and undirected conditions. Mean linearity scores were 1.00 ± 0.00 (SD) for directed songs and 1.00 ± 0.00 for undirected songs from the same birds (Table 2), and mean consistency scores were 0.91 and 0.92, respectively (Table 2; P = 0.73; paired sign test). The difference between our findings and those of Sossinka and Böhner is likely to reflect a difference in the degree of stereotypy in the undirected songs of birds in the two studies. In our study, the undirected songs of adult zebra finches were very stereotyped, with high linearity and consistency scores similar to those of other recent studies. For example, Zevin et al. (2004) reported a mean linearity score of 0.98 and a mean consistency score of 0.96 for the undirected songs of 27 adult zebra finches. This high level of sequence stereotypy presents little opportunity for songs to become even more stereotyped in the directed condition. In contrast, for the one bird for which data are available in the study by Sossinka and Böhner (1980), the linearity score of undirected song can be calculated as 0.64 (insufficient data are available to calculate sequence consistency). This value is lower than that for any of the adult zebra finches to which this measure was applied in our study (n = 10) or in the study by Zevin et al. (2004; n = 27). Because the birds from the study of Sossinka and Böhner were a combination of wild caught and domesticated birds of unknown ages, we speculate that the greater variability in the sequence of undirected songs in their study reflects either age or strain differences. With respect to the former possibility, it will likely be informative to test the social modulation of song stereotypy in a population of juvenile birds before song crystallization.

View this table:
TABLE 2.

Analysis of song sequence

Because we found no evidence for the social modulation of sequence stereotypy, we did not have a strong expectation that lesions of LMAN would affect this parameter of song. However, in juvenile birds, lesions and inactivation of LMAN have been reported to increase the stereotypy of syllable sequencing (Bottjer et al. 1984; Ölveczky et al. 2005; Scharff and Nottebohm 1991). Hence, we measured sequence linearity and consistency for directed and undirected songs from birds 1 wk postlesion and compared these values to those from the same birds prelesions (Table 2). We found no influence of LMAN lesions on these measures for either directed or undirected songs (linearity: P > 0.99, 1 sample sign test; consistency: P > 0.99, paired sign test). Our finding that LMAN lesions do not increase sequence stereotypy in adult birds may again result from a ceiling effect: the prelesion songs in our study were already sufficiently stereotyped to make it difficult to detect any further increases.

Although song linearity and consistency did not differ significantly across social context, we did notice a difference between these conditions in the stereotypy with which birds terminated their songs. In particular, there were often more types of song terminations in directed songs than in undirected songs. We quantified this by calculating the percentage of syllables within a motif that were observed to precede song terminations. Across the population of normal adult birds, there were significantly more types of song terminations in directed versus undirected songs (Table 2; 9/10 birds, P = 0.02, paired sign test). This difference, which reflects greater variability in directed versus undirected songs, may result from the less predictable, and potentially disruptive, auditory and visual stimulation from the female bird that is present in the directed context. Indeed, it has been reported previously that unexpected auditory and visual stimuli can elicit premature song terminations Cynx 1990; Cynx and Von Rad 2001). Regardless of the source, we found that this difference between directed and undirected song persisted in all birds after lesions of LMAN (Table 2).

SONG TEMPO.

To characterize context-dependent changes to song tempo, we measured, for each bird, the durations of complete motifs produced in directed and undirected songs. Figure 8A shows data from one bird for which there was a robust modulation of song tempo by social context. In this case, song was 2.6% slower in the undirected condition than in the directed condition (P < 0.0001; Mann-Whitney U test). The degree of tempo modulation by social context varied across birds and across recording sessions for individuals. Across all control recordings of interleaved directed and undirected songs (3 recording sessions each for 5 birds prelesions, 2 birds presham lesions, and 3 control birds), song was faster in the directed context for 24/30 recording sessions. This difference was significant for 20 of these sessions (P < 0.05; Mann-Whitney U test for comparisons within 1 recording session for each bird). For 1 of 30 sessions, song was significantly slower in the directed context (P < 0.02; Mann-Whitney U test). Overall, there was a highly significant effect of social context on song tempo, with an average slowing of song by 1% in the undirected context (P < 0.0001; paired sign test, n = 30 control recordings; P = 0.02, 1 sample sign test for average effect per bird, n = 10 birds). This finding is in accordance with that of Sossinka and Böhner (1980), who reported an average slowing of undirected song by ∼2.7% relative to directed song and significant context-dependent tempo differences in six of eight birds.

FIG. 8.

Context-dependent differences in song tempo are dependent on input from LMAN. A: top: DS is typically produced at a faster tempo than US. Spectrograms of a representative motif during US (top) and DS (bottom) of 1 bird are aligned by onset of the 1st syllable. Dotted lines indicate onset time and offset times of motifs. Bottom: histogram of motif durations during US (open bars, n = 37) and DS (green bars, n = 35). Motif duration was significantly shorter during DS (means: 640.6 vs. 657.2 ms; P < 0.0001, Mann-Whitney U test). B: lesions of LMAN eliminate context-dependent differences in song tempo. Undirected motifs were significantly slower than directed motifs in 4 of 5 birds before lesions of LMAN (P < 0.01, Mann-Whitney U test for comparisons within 1 recording session for each bird). After lesions of LMAN, song tempo was not significantly different in 4 of 5 birds (P > 0.14, Mann-Whitney U test). In the remaining bird, undirected song was significantly faster than directed song (P < 0.04, Mann-Whitney U test). Lines connect data points from the same bird. Bars indicate group means; error bars indicate SE.

To assess whether the modulation of song tempo by social context is dependent on inputs from LMAN, we compared the magnitude of tempo modulation for recordings 1 day before lesions and 1 wk after lesions (Fig. 8B). In all cases, before lesions, the tempo was faster in the directed context than the undirected context. This difference was significant for four of five birds considered individually (P < 0.01, Mann-Whitney U test). In contrast, 1 wk after lesions of LMAN, context-dependent differences in song tempo were no longer consistent: four of five birds did not exhibit a significant difference in tempo between directed and undirected contexts (P > 0.05, Mann-Whitney U test for recordings in each bird), and in the remaining bird, undirected song was significantly faster than directed song (P < 0.01, Mann-Whitney U test). Therefore lesions of LMAN abolished the social modulation of song tempo.

Stabilization of syllable structure with age

Previous studies in adult zebra finches have shown that song plasticity in response to manipulation of auditory feedback declines with age (Brainard and Doupe 2001; Lombardino and Nottebohm 2000). Because variability in motor output is an important component of feedback-based motor learning, we asked whether the decline in plasticity is accompanied by an age-dependent decline in song variability. Such a decline is suggested by the gradual reduction in the variability of undirected song for control and sham-lesioned birds over the 8- to 10-wk period of recordings (Fig. 5A). To explicitly examine this issue, we recorded directed and undirected songs for 17 additional adult zebra finches. For these birds, we indeed found that socially driven modulation of song variability is attenuated with age. In young adults (<6 mo old), variability in FF was significantly greater during undirected song than during directed song (Fig. 9A, left; 28/29 syllables in 11 birds; P < 0.0001, paired sign test). These differences were significant for 19/29 syllables (F-test for equality of variance across conditions for each syllable). In contrast, variability in FF did not differ across social context in birds >4 yr old (Fig. 9A, right; n = 13 syllables in 6 birds; P = 0.5811, paired sign test). The lack of context-dependent differences in syllable variability in older birds seemed to be the result of an age-dependent reduction in the absolute level of variability in FF during undirected song to the level present during directed song (Fig. 9A). That is, variability in FF during undirected song was significantly lower in older birds than in young adults (Fig. 9A; P = 0.0015, Mann-Whitney U test), reminiscent of the lesion-induced stabilization of syllable structure (cf. Figs. 5B and 9A).

FIG. 9.

Age-dependent stabilization of syllable structure despite persistent context-dependent differences in LMAN activity. A: context-dependent differences in song variability decline with age. CV of FF is plotted for syllables recorded in both behavioral contexts (1–7 syllables/bird); lines connect data points for the same syllable. Bars indicate group means. Variability in FF was significantly greater during US than during DS in birds <6 mo old (left: 28/29 syllables in 11 birds; P < 0.0001, paired sign test) but not in older birds (right: >4 yr old; 13 syllables in 6 birds; P = 0.5811, paired sign test). Variability in FF during US was significantly lower in older adults compared with young adults (P = 0.0015, Mann-Whitney U test). B: context-dependent differences in LMAN multiunit activity in an adult male (>5 yr old). Top: singing-related activity in LMAN during 3 renditions of the motif during US (left) and DS (right). Middle: rectified, smooth waveforms of LMAN multiunit activity during 10 renditions of motif are superimposed for each behavioral condition. Waveforms are normalized by mean spontaneous activity (dotted line). LMAN activity was greater in magnitude and more variable in pattern across repeated renditions during US. Bottom: CV of activity level was calculated in 1-ms bins across repeated undirected (left) and directed (right) motifs (thick black lines). CV across undirected motifs (gray) is replotted for comparison with CV across directed motifs (right). Arrows denote the mean time-averaged CV in each condition. C: context-dependent differences in LMAN activity are present in birds across a range of ages. Time-averaged CV during US and DS are plotted for each site. Data in B are indicated by closed squares. Variability in LMAN activity was significantly greater during US than during DS in birds of different ages (left: <6 mo old: 10/10 sites in 5 birds; P = 0.002, paired sign test; right: >4 yr old: 9/9 sites in 2 birds; P = 0.004, paired sign test). Although context-dependent differences were present in older birds, the absolute level of variability in LMAN activity declined significantly in both behavioral contexts with age (undirected CV: 0.155 vs. 0.123; P = 0.01, Mann-Whitney U test; directed CV: 0.087 vs. 0.068; P = 0.004, Mann-Whitney U test).

As with LMAN lesioned birds, the age-dependent loss of socially driven modulation of FF did not result from a general disruption of courtship behavior. Older adult males exhibited robust context-dependent differences in higher-level features of song. In six of six birds, the number of introductory elements was significantly greater during directed song than during undirected song (Fig. 7F; P < 0.05, Mann-Whitney U test for comparisons within 1 recording session for each bird), and in four of six birds, the number of motifs per bout was significantly greater during directed song (Fig. 7G; P < 0.05, Mann-Whitney U test). Thus the social modulation of higher-level features of song persisted despite the absence of context-dependent differences in syllable structure, consistent with the hypothesis that different parameters of song are independently controlled.

The stabilization of syllable structure induced by age and by lesions of LMAN raises the question of whether or not the mechanism for song stabilization is the same in the two groups of birds. The observed age-dependent stabilization of syllable structure may reflect a normal developmental decline in factors from the AFP that promote variability. For example, variability in the pattern of AFP activity may decline with age or may not differ across social context in older adult birds. Alternatively, signals from the AFP may not change with age. Rather, as synaptic connections in RA are strengthened, the ability of extrinsic signals to modulate ongoing motor patterns and subsequent song output may decline with age (Lombardino and Nottebohm 2000).

To examine these possibilities, we compared the singing-related activity in LMAN in birds over a wide range of ages. Figure 9B shows the singing-related activity during one recording session for a male bird >5 yr old. Across repeated renditions of the bird’s motif, the pattern of activity was more reproducible during directed song than during undirected song. Consistent with previous reports, we found that variability in the pattern of multiunit activity in LMAN was significantly greater during undirected song than during directed song in young adult birds (<6 mo old; Fig. 9C, left; 10/10 sites in 5 birds; P = 0.002, paired sign test). Similarly, in older adults (>4 yr old), variability in LMAN activity was also greater during undirected song (Fig. 9C, right; 9/9 sites in 2 birds; P = 0.004, paired sign test). Moreover, the degree to which variability in LMAN activity differed across social context was similar between birds of different ages (mean ratio of the undirected neural CV to the directed neural CV in young and older adults: 1.78 and 1.81, respectively).

However, despite the presence of context-dependent differences in LMAN activity in birds of different ages, we found that the overall level of variability in LMAN activity was significantly lower in older adults than in young adults (Fig. 9C). This age-dependent decline in the absolute level of LMAN variability was apparent in both behavioral contexts (Fig. 9C; undirected CV 0.155 vs. 0.123; P = 0.01, Mann-Whitney U test; directed CV: 0.087 vs. 0.068; P = 0.004, Mann-Whitney U test).

Together, these findings suggest that multiple factors contribute to the age-dependent stabilization of syllable structure: 1) the susceptibility of the motor pathway to sources of perturbation may decline with age and 2) the factors that promote variability in motor activity, such as variability in LMAN activity, may decline with age.

DISCUSSION

Contributions of LMAN to context-dependent regulation of song and behavior

In zebra finches, adult song and behavior are strongly modulated by social context. During directed songs produced in a courtship context, the number of introductory notes, the number of song motifs, and song tempo are increased, and the variability of syllable structure is decreased relative to that of undirected songs produced by birds in isolation (Dunn and Zann 1996; Kao et al. 2005; Sossinka and Böhner 1980). Moreover, directed songs are accompanied by a courtship dance that is not produced when the birds are alone (Williams 2001; Zann 1996). This switching of song and behavioral state occurs within the first moments after introduction of a female to the presence of a male and reverses equally rapidly after her removal (Hessler and Doupe 1999b).

Several previous observations have supported the possibility that signals from the AFP contribute to this rapid, context-dependent modulation of song and behavior: 1) neural activity in the AFP is greater and more variable under conditions of undirected song than directed song, indicating that modulation of neural activity in the AFP is correlated with modulation of behavior (Hessler and Doupe 1999b; Jarvis et al. 1998; Kao et al. 2005) and 2) active manipulation of signals from the AFP, by microstimulation of LMAN during singing, can alter the acoustic structure of song elements, indicating that signals from the AFP are sufficient to drive changes in song in real time (Kao et al. 2005). Here, we showed that the nucleus LMAN is necessary for a specific subset of the context-dependent changes in song and behavior. We show that, in normal adult birds, there is a robust decrease in the variability of syllable structure in directed (courtship) song versus undirected song and that this difference in song variability is eliminated by lesions of LMAN. In addition, lesions of LMAN eliminate the social modulation of song tempo. In contrast, lesions do not prevent context-dependent changes in other features of song, including the number of introductory elements that precede song and the number of motifs that are produced during a bout of singing. Similarly, lesions of LMAN do not prevent the postural changes and courtship dance that normally accompany directed song.

The selectivity of the effects of LMAN lesions has two major implications. First, the effects of lesions on song modulation are not the consequence of globally disrupting the male bird’s ability to detect and respond to social cues. When females are presented, the courtship vigor of lesioned males is as robust as that of control birds. Moreover, lesions of LMAN did not affect the social modulation of syntactic features of song, such as the number of introductory elements and the number of motifs per bout. These results indicate that LMAN is not an obligatory structure for integrating and representing information about social context. Rather, LMAN must be downstream from brain regions that register such information.

Second, the selectivity of the effects of lesions of LMAN indicates that separate brain regions participate in the modulation of different components of song and courtship behavior. This finding is consistent with the anatomy of the song system: neurons in LMAN project directly to the premotor nucleus RA (Fig. 1A), which is thought to provide motor commands that control the precise structure of individual song elements (Vu et al. 1994; Yu and Margoliash 1996). Moreover, microstimulation in LMAN at low intensities modulates syllable structure of adult birds without affecting syllable sequencing (Kao et al. 2005; Vu et al. 1994). In contrast, other aspects of song structure, such as patterning of introductory elements and motifs, are thought to be controlled by nuclei earlier in the premotor pathway for song, such as HVC and its inputs (Foster and Bottjer 2001; Vu et al. 1994).

Recent measurements of the social modulation of song tempo have shown that expiratory pulses, which contribute to syllable production, are longer during undirected song than during directed song, whereas inspiratory pulses are largely unchanged across social context (Cooper and Goller 2006). These observations are consistent with the social modulation of song tempo reported by Sossinka and Böhner (1980) and with the observation by Glaze and Troyer (2006) that durations of syllables can be regulated independently from durations of intervening intervals. These data suggest that increases in song duration may arise from an accumulation of increases in the durations of individual song elements. In this regard, it is possible that the observed effect of lesions of LMAN on context-dependent differences in song tempo reflects the direct modulation by LMAN of pattern generation circuitry in RA that contributes to the structure of individual syllables. Alternatively, or in addition, lesions of LMAN might affect song tempo by indirectly influencing activity in other song premotor areas, such as HVC, which are ultimately modulated by activity in LMAN and RA through recurrent connections (Ashmore et al. 2005; Vu et al. 1998).

Regardless of the mechanism, the notion that LMAN can modulate song tempo is consistent with previous reports that lesions of LMAN result in a gradual acceleration of the tempo of undirected songs (Brainard and Doupe 2001; Williams and Mehta 1999). Although such increases in tempo occur over a period of many days to weeks, in contrast to the rapid modulation of song tempo by social context, they are nonetheless consistent with our observation that LMAN can influence song tempo.

Possible functions of different song types (practice versus performance)

For song learning, variability in motor output may be required for evaluation mechanisms to differentially reinforce those patterns of motor activity that produce better songs (i.e., closer to desired targets) and/or to punish the motor commands that result in worse songs. After song is learned, the ability to modify it likely remains important to compensate for age-related perturbations in the motor program and song structure. Such perturbations may arise from changes in hormone levels, changes in the peripheral motor structures for song production (e.g., muscle tone and innervation), and/or the birth, death, and incorporation of new neurons in the motor pathway (Alvarez-Buylla 1990; Nordeen and Nordeen 1988; Scharff 2000).

Given the importance of variability in motor learning, it is quite striking that behavioral variability between directed and undirected conditions is robustly modulated. While the reasons for this modulation of song variability remain unclear, one intriguing possibility is that undirected song is a state of motor practice in which male birds try out alternative motor commands, produce a range of vocalizations, and use auditory feedback to optimize and/or maintain the song. In contrast, directed song may reflect a state of motor performance in which individuals exploit what they have already learned to select the patterns of motor activity that produce their “best” current version of song (Jarvis et al. 1998; Kao et al. 2005; Sutton and Barto 1998).

These considerations render it plausible that the AFP is part of nervous system circuitry that specifically contributes to the generation of motor variability for the purposes of learning (Bottjer 2004; Brainard 2004; Doya and Sejnowski 2000; Kao et al. 2005; Ölveczky et al. 2005). In principle, the reduced variability of directed song could reflect an active process in which the nervous system tightly controls activity in the motor pathway to produce optimal performance, whereas the greater variability of undirected song could reflect a noisier default motor performance in the absence of an audience. However, our data show that removal of inputs from the AFP to the motor pathway by lesions of LMAN causes a reduction in the variability in syllable structure present during undirected song to the level present during directed song (Figs. 4B and 5B). These results support the view that the increased variability present in undirected song is actively introduced by the AFP. They are therefore consistent with the hypothesis that one of the critical contributions of the AFP to vocal plasticity may rely on its ability to introduce variability to the song motor pathway to generate motor exploration during undirected song. A further implication is that the AFP, and perhaps basal ganglia circuits in other species, may contribute to an active switching of motor activity and behavior between states of practice and performance.

Previous work has shown that lesions of LMAN can prevent experience-dependent changes to adult song (Brainard and Doupe 2000; Morrison and Nottebohm 1993; Williams and Mehta 1999), suggesting two possible roles for LMAN: 1) LMAN may provide experience-dependent instructive signals that actively drive changes in the song motor pathway to produce a good match to the tutor song or 2) LMAN may provide permissive factors for song plasticity, without which song cannot change, but which by themselves need not provide the direction for change (for discussion, see Brainard 2004). In principle, variability introduced into adult song by the AFP could be construed as a permissive factor for plasticity: the songs of adult birds with LMAN lesions may remain stable because motor exploration is effectively prevented. However, the finding that LMAN can direct moment-by-moment changes in song also indicates that signals from the AFP have the potential to actively instruct song plasticity, perhaps by persistently biasing motor output toward desired targets (Kao et al. 2005; Troyer and Doupe 2000).

Stabilization of song by age and by lesions of the AFP

If variability of syllable structure contributes to motor exploration and song plasticity, one might expect to see changes in the degree of variability under natural conditions associated with song stabilization. One such condition is aging, which in adult zebra finches (beyond the normal age of crystallization at ∼90 days) leads to a progressive stabilization of song. This stabilization has been documented as a decline in the effects of removing auditory feedback as birds age (Brainard and Doupe 2001; Lombardino and Nottebohm 2000). Here, we show that the age-dependent stabilization of song is indeed accompanied by changes in the variability of syllable structure. Although there was no apparent change in the variability of syllables produced by older males during directed singing, the variability of syllables produced during undirected singing dropped to the level present in directed song (Fig. 9A). Hence aging and LMAN lesions had remarkably similar effects on both the level of syllable variability and its modulation by social context.

This similarity between the effects of aging and LMAN lesions on syllable variability raises the question of whether LMAN is functionally inactive in older birds. We found that there was indeed a decline in variability of LMAN neural activity between younger and older birds (Fig. 9C). However, even in the oldest birds (>4 yr), where social modulation of song variability was no longer present, there continued to be social modulation of variability of neural activity within LMAN (Fig. 9C). This suggests that enhanced stability of adult song derives from multiple processes, including 1) reduced variability in the patterns of neural activity present within LMAN and 2) reduced sensitivity of the motor pathway to such activity. Such reduced sensitivity could derive from a progressive weakening of inputs from LMAN to RA and/or a progressive strengthening of connections within the motor pathway itself, such that it becomes less influenced by inputs from LMAN.

The finding that the modulation of song variability declines with age despite the persistent modulation of LMAN activity suggests that factors in addition to patterned neural activity from LMAN are likely to contribute to regulation of song variability. These factors may include slower time scale processes, such as neural or trophic influences on synapse formation and the survival and incorporation of new neurons in the motor pathway (Johnson and Bottjer 1994; Kittelberger and Mooney 1999, 2005; Scharff and Nottebohm 1991). Indeed, previous studies have shown that LMAN itself provides trophic support to RA during development (Akutagawa and Konishi 1994; Johnson and Bottjer 1994). Therefore LMAN may well influence both rapid (synaptic) and slower (trophic) time scale processes that contribute to the physiology of the motor pathway and the structure and plasticity of song.

GRANTS

This work was supported by a Howard Hughes Medical Institute (HHMI) Predoctoral Fellowship to M. H. Kao; the MacArthur Foundation, National Alliance for Research on Schizophrenia and Depression, the Sandler Family Supporting Foundation, the Steven and Michele Kirsch Foundation, and National Institute of Mental Health Grant MH-55987 to A. J. Doupe; and the McKnight Foundation, the Klingenstein Fund, the HHMI Biomedical Research Support Program grant, a Searle Scholars Award, a Sloan Research Fellowship, and National Institute of Deafness and Communication Disorders Grant R01 DC-006636-01 to M. S. Brainard.

Acknowledgments

We thank A. Arteseros, L. Bocksai, and K. McManaway for technical assistance, E. Ihle and S. C. Woolley for behavioral scoring, and A. J. Doupe for thoughtful comments on this manuscript.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract