Norepinephrine in the avian auditory cortex enhances developmental song learning

Sensory learning during critical periods in development has lasting effects on behavior. Neuromodulators like dopamine and norepinephrine (NE) have been implicated in various forms of sensory learning, but little is known about their contribution to sensory learning during critical periods. Songbirds like the zebra ﬁ nch communicate with each other using vocal signals (e.g., songs) that are learned during a critical period in development, and the ﬁ rst crucial step in song learning is memorizing the sound of an adult conspeci ﬁ c ’ s (tutor ’ s) song. Here, we analyzed the extent to which NE modulates the auditory learning of a tutor ’ s song and the ﬁ delity of song imitation. Speci ﬁ cally, we paired infusions of NE or vehicle into the caudomedial nidopallium (NCM) with brief epochs of song tutoring. We analyzed the effect of NE in juvenile zebra ﬁ nches that had or had not previously been exposed to song. Regardless of previous exposure to song, juveniles that received NE infusions into NCM during song tutoring produced songs that were more acoustically similar to the tutor song and that incorporated more elements of the tutor song than juveniles with control infusions. These data support the notion that NE can regulate the formation of sensory memories that shape the development of vocal behaviors that are used throughout an organism ’ s life. NEW & NOTEWORTHY Although norepinephrine (NE) has been implicated in various forms of sensory learning, little is known about its contribution to sensory learning during critical periods in development. We reveal that pairing infusions of NE into the avian secondary auditory cortex with brief epochs of song tutoring signi ﬁ cantly enhances auditory learning during the critical period for vocal learning. These data highlight the lasting impact of NE on sensory systems, cognition, and behavior.


INTRODUCTION
The processing and memorization of sensory information are fundamental to behavioral plasticity (i.e., learning). For example, learning the sensory attributes of others is important for individual recognition and social decision-making, and learning the sensory attributes of food and predators (as well as the vocalizations associated with food and predators) are critical for survival (1)(2)(3)(4). Sensory learning can occur at all life stages, but sensory learning during critical periods in development is particularly impactful because it leads to persistent sensory representations that shape behavior throughout an organism's life (5,6). As such, revealing the mechanisms that modulate sensory learning during critical periods in development is fundamental to our understanding of behavioral and neural plasticity (reviewed in Refs. 7 and 8).
Songbirds provide a powerful opportunity to discover neural mechanisms that modulate sensory learning during critical periods in development (9)(10)(11)(12)(13)(14). Like developing humans, juvenile songbirds learn their songs by first memorizing the song of an adult tutor ("sensory learning") during a critical period in development, and then engaging in a protracted period of vocal practice and sensorimotor learning (e.g., using sensory feedback to shape the development of their song; Refs. 10 and 15-17). Because sensorimotor song learning involves comparing current vocal performance to a sensory representation of the target song, sensory learning serves as the foundation for the sensorimotor learning of song (18)(19)(20)(21)(22). Furthermore, sensorimotor learning occurs over weeks to months, progresses even without the continued presence of a tutor, and can start well after the sensory period closes; consequently, accurate sensory memories formed during sensory learning must persist in order for songbirds to develop accurate imitations of the tutor song (reviewed in Ref. 17). Given the similarities between mechanisms of developmental song learning in songbirds and speech acquisition in humans (including a reliance on sensory learning) and homologies in auditory circuitry between mammals and songbirds (18,23), analyses of birdsong learning have important implications for understanding mechanisms of speech acquisition in humans.
Numerous studies have highlighted the importance of the caudomedial nidopallium (NCM), a higher order auditory processing area, in sensory learning in songbirds (12,13,(24)(25)(26)(27)(28)(29)(30)(31)(32)(33). For example, song learning leads to a subset of NCM neurons becoming selectively tuned to the tutor song (34), and manipulations of signaling cascades in NCM during developmental song learning affect tutor song imitation (35,36). However, relatively little is known about the factors that modulate NCM activity to influence sensory learning in the service of developmental song learning. For example, no study to date has analyzed how neuromodulators in NCM affect the sensory learning of a tutor's song during the critical period for song learning.
Catecholamines such as norepinephrine (NE) have been found to modulate neural plasticity and learning (37)(38)(39)(40)(41)(42). For example, pairing playback of tones with optogenetic stimulation of the terminals of NE neurons that project to the primary auditory cortex leads to enhanced behavioral discrimination of those tones and to greater plasticity in the cortical representation of sounds in rodents (43,44). Norepinephrine has also been implicated in sensory learning in adult songbirds. Neurons in the locus coeruleus (LC) innervate the NCM and are active during auditory learning (30,45,46). Further, NCM neurons are replete with NE receptors, and antagonizing NE receptors in NCM reduces neural manifestations of auditory learning in adult songbirds (30,45,47). Although studies of NE contributions to sensory processing and learning have primarily been conducted in adult songbirds, there is some evidence suggesting that NE neurons could be important for the sensory learning required for developmental song learning; specifically, tutoring paradigms that enhance song learning are associated with more noradrenergic LC neurons expressing immediate early genes (46). However, it is unknown whether NE acting within auditory processing areas modulates the sensory learning central to developmental song learning. As such, we analyzed how manipulations of NE in the NCM during developmental song tutoring modulated the strength and fidelity of song learning.

Animals
Male juvenile zebra finches (Taeniopygia guttata) were bred and raised at McGill University (n = 21 juveniles from 11 nests). Birds were kept on a 14 L:10 D photoperiod, with food and water provided ad libitum. Procedures were performed according to protocols approved by the McGill University Animal Care and Use Committee and were in accordance with guidelines from the Canadian Council on Animal Care and Use Committee.
Juveniles for song tutoring experiments ("pupils") were born in a colony at McGill University. "Song-naïve juveniles" (n = 13 juveniles from 6 nests) were raised with both parents until 5-7 days of age, at which time fathers were removed. Because there is minimal song learning before 20 days posthatching (dph), such exposure to fathers should not affect song learning (48,49). After reaching nutritional independence ($40 dph), juveniles were housed individually in a sound-attenuating chamber ("soundbox") for a few days before surgery and experimental tutoring. Tutoring began when juveniles were 52 ± 1 (range: 45-60) dph, and the age at which tutoring started was not significantly different across experimental groups.
In addition to evaluating song learning in song-naïve birds, we assessed song learning in "song-experienced juveniles" (n = 8 juveniles from 5 nests); these birds were housed with both their father and mother until $40 dph and thereafter housed individually in a soundbox. Previous experiments have demonstrated that even with this limited exposure to their father's song, song-experienced birds produce songs that resemble their father's song (48)(49)(50). Experimental tutoring began when juveniles were 54 ± 0 (range: 54-56) dph, with ages not significantly different between experimental groups and from song-naïve juveniles.

Cannulae Implantation
For surgical cannula implantation, birds were anesthetized with an intramuscular injection of ketamine (0.02 mg/ g; IM) and midazolam (0.0015 mg/g; IM) followed by gaseous isoflurane (0.2%-4.0%; O 2 flow 1.5-2 L/min). Thereafter, birds were placed in a stereotaxic device, with their beak stabilized at 45-degrees. Guide cannulae with dummy probes (Harvard Apparatus, Hollison, MA, CMA 7) were bilaterally targeted at NCM (0.5 mm rostral from the caudal edge of the bifurcation of the midsagittal sinus, 0.5 mm lateral from the midline, 2.5 mm deep from the surface of the brain). Cannulae were then fixed to the bird's skull with epoxy and dental cement. The cannulae remained fixed and in place for the duration of this study for all animals, and there were no signs of brain lesions in histology.

Drug Infusions and Song Tutoring
After 4-7 days of recovery from surgery, dummy probes in the guide cannulae were replaced with microdialysis probes (Harvard Apparatus, Hollison, MA, CMA 7). CMA 7 microdialysis probes were modified into an untethered reservoir system to allow for passive diffusion of drug in freely behaving birds (51)(52)(53)(54). Specifically, inlet and outlet tubes were cut to 4-6 cm in length and served as a reservoir for fluids. Twice each day, both probes were slowly filled (10 mL/min) with 40 mL of 0.025 M phosphate-buffered saline (PBS; pH = 7.4) using a syringe pump (Harvard Apparatus, Pump 11 Elite). Inlet and outlet tubes were outfitted with caps to minimize fluid evaporation. Juvenile songs were recorded for 1-2 days before the onset of tutoring.
Experimental tutoring started 1-3 days following the insertion of microdialysis probes. On the days of experimental tutoring, probes were slowly filled with either PBS or 1 mM norepinephrine (NE; Sigma A7256; dissolved in sterile PBS) using the same infusion parameters as on previous days. This dose is within the range of doses previously used in songbirds to study the effects of NE on neural activity (0.5-20 mM; Refs. 32, 55, and 56). Fifteen minutes after infusion with PBS or NE, pupils were tutored with song playback of an adult zebra finch ("tutor"). Contents of the probe were washed out with PBS 15 min after the termination of each tutoring session (Fig. 1).
Pupils were tutored on 5 consecutive days with playbacks of a tutor's song. Tutoring stimuli consisted of 8-10 different renditions of an adult male zebra finch's song (tutor song) played in a pseudorandom order. Each tutoring session lasted 30 min and consisted of playbacks of 30 song renditions (1 song/min), with the same set of songs across all 5 days of experimental tutoring. Song-naïve juveniles were tutored with songs from one of two adult zebra finches (i.e., two sets of tutor stimuli but each juvenile only heard one of the sets). The songs of the tutors were acoustically distinct and rated to be only 26% similar to each other (see below for quantification). Tutor songs differed in motif durations and the numbers of syllables in their motifs (tutor 1's motif was 1,097 ± 8 ms in duration and contained 7 syllables; tutor 2's motif was 736 ± 8 ms in duration and contained 5 syllables). Song-experienced juveniles (i.e., sequential tutoring experiments) were tutored with songs from one of these tutors (tutor 1). For all experiments, the songs of the tutor were screened to be acoustically distinct from that of the juvenile's biological father.
Probes were removed 2 days after the last day of tutoring and were replaced with dummy probes. Juveniles remained individually housed until adulthood (4 mo) to allow for song development without the influence of other birds (14,46,57,58).
To try to control for potential genetic contributions to song learning, we separated juveniles from the same nest across experimental groups when possible. Song-naïve juveniles were hatched from six different nests, three of which produced clutches of multiple males, and siblings from two of these three clutches were allocated to different experimental groups (Supplemental Table S1; all Supplemental material is available at https://github.com/ sakatasongbirdlab/Chen_Sakata_NE_NCM; https://doi.org/ 10.5281/zenodo.4750432). Song-experienced males were hatched from five different nests, two of which produced clutches of multiple males, and males from both these clutches were allocated to different experimental groups (Supplemental Table S1).

Song Recording and Analysis
The adult songs of all pupils were recorded when birds were 4 mo old using an omnidirectional microphone (Countryman Associates, Menlo Park, CA) positioned above the bird's cage [Sound Analysis Pro (SAP) v.2011.10; digitized at 44.1 kHz; http://ofer.sci.ccny.cuny.edu/sap/sound-analysis-pro]. All song exemplars of pupils and tutors were spontaneous songs produced in isolation ("undirected song").
The degree to which the motifs of a pupil's adult song resembled the motifs of their tutors or fathers was assessed using both quantitative and qualitative (human) ratings of similarity (46,59). For computerized calculations of song similarity, we randomly selected 30 pupil song bouts within a day and extracted the first motif of each bout (introductory notes were excluded). A motif is a stereotyped sequence of syllables that is repeated throughout the song bout. Using the "Explore and Score" function in SAP, we compared each of these pupil motifs to three motif renditions of their tutor or father. We analyzed the similarity score obtained from SAP, which, broadly speaking, reflects the percent of sounds in the tutor's song that are observed in the pupil's song (asymmetrical comparison; Ref. 60). We computed the average similarity scores across the 30 renditions of a pupil's motif to each of the tutor or father's exemplars and then calculated the median SAP similarity score across the exemplars.
To complement quantitative calculations of similarity, we also collected human ratings of acoustic similarity (46,59). For this, three researchers with extensive experience analyzing zebra finch song and who were blind to experimental conditions rated the similarity between two exemplars of a pupil's motifs and one exemplar of the tutor or father's motif. Like the algorithm implemented in SAP to measure song similarity, human raters were instructed to score similarity based on the percent of sounds in the tutor or father's song that were observed in the song of the pupil. Scores were averaged across raters and renditions to compute a human rating of similarity for each pupil.
For the analysis of experimental effects on song learning, we averaged SAP and human scores to obtain a single metric of similarity to tutor ("song similarity"; Ref. 46). Human and SAP similarity scores were significantly correlated (Supplemental Fig. S1), and the effect of experimental manipulations was comparable when SAP and human scores were independently analyzed.
To gain further insight into song learning, we asked the same three raters to classify individual syllables in the pupil's songs as belonging to the tutor and/or father's song, depending on the experiment (see RESULTS). Raters were also allowed to assign syllables as not belonging to either the tutor or father's song (i.e., as "novel syllables"). They were given syllables from three renditions of pupil motifs along with several "distractor" syllables (i.e., not from father or tutor), and syllable assignment was done blind to experimental condition, syllable context, and whether template syllables belonged to the father or the tutor. All of the distractor syllables were correctly classified as not belonging to the father or tutor. Raters unanimously classified >70% of syllables, and Fleiss' kappa for classification scores ranged from 0.57-0.60, suggesting moderate to substantial agreement among raters (61,62). The final quantification of the number of syllables learned were based on the averages across the three raters and were used to compute and analyze the number of a tutor or father's syllables that were incorporated into the pupil's song.
To assess whether tutoring led to significant learning, song similarity of experimentally tutored birds were compared to that of adult birds deprived of exposure to a tutor song ("untutored birds"; n = 13). These birds were raised like song-naïve juveniles but were housed individually from nutritional independence and not exposed to the song of adult male zebra finches throughout development. Songs from three untutored birds from McGill and 10 untutored birds from the University of California, San Francisco were used in this analysis (46,63). Song similarity scores were computed between two motifs from each untutored bird and a single motif of a tutor's song. We computed these song similarity scores for each untutored bird against each individual tutor or father and calculated the average similarity scores across untutored birds for each tutor or father. These similarity scores reflect the degree to which songs of untutored birds could, by chance, resemble the song of a tutor or father and were used to assess the significance of learning (see RESULTS).

Histology
After recording their adult songs, birds were perfused to evaluate the position of the cannulae when they were 4-5 mo old. For this, birds were given an overdose of gaseous isoflurane and then transcardially perfused with 25 mL of heparinized saline (100 IU/mL) followed by 150 mL of 4% paraformaldehyde (pH = 7.4). Brains were removed, postfixed overnight at 4 C in 4% paraformaldehyde, and then transferred to a 30% sucrose solution overnight at 4 C for cryoprotection. Sagittal sections were cut at 40 μm using a sliding microtome (Leica Biosystems, Wetzlar, Germany) and then stored in 0.025 M PBS containing 0.05% sodium azide at 4 C. Every third section from each hemisphere of each animal was mounted and stained using cresyl violet to visualize cannulae locations.
Verification of cannulae location was done blind to the experimental treatment. For each hemisphere of each bird, an experimenter first noted all sagittal sections where the cannula was observed. From this handful of sections, the median section was selected as the reference section to measure cannula placement. A reference line was drawn from the most dorsal aspect of the striatum to the most caudal portion the nidopallium (close to the most ventral tip of the hippocampus) on the reference section (see Supplemental Fig. S2). The rostral-caudal extent of the cannula was measured along a line parallel to the reference line from the ventral tip of the cannula to the caudal edge of the nidopallium. The depth of the probe was measured as the distance from the ventral tip of the cannula to the dorsal edge of the nidopallium along a line perpendicular to the reference line. We measured the depth of the probe from the dorsal edge of the nidopallium and not the dorsal edge of the hippocampus because the hippocampus was damaged or lost in the histological sections of some birds. The medial-lateral placement of the cannula was computed based on the number of histological sections lateral to the midline.
NCM is located in the caudal part of the nidopallium, generally within 1,000 mm from the midline. Although neural activity in response to sounds are observed throughout the NCM region, the caudal part of the NCM has been found to be particularly important for song learning (reviewed in Ref. 23). We classified cannula that were positioned <1,000 mm from the midline, <650 mm rostral to the caudal border of the nidopallium, and <1,500 mm deep as being positioned within NCM ("hits"). Based on this criterion, all but three birds were considered to have cannulae bilaterally placed within NCM ("bilateral hits"). Birds with "bilateral hits" had cannula centered <1,000 mm from the midline, 405 ± 19 (mean ± SE) mm rostral to the caudal border of the nidopallium, and 1,005 ± 21 mm below the dorsal edge of the nidopallium. One bird was categorized as having a "unilateral hit," and two birds were considered to have both of their cannula outside of the NCM ("bilateral misses"; Supplemental Table  S1). Although the mediolateral and dorsoventral coordinates of these cannula were similar to those of other birds, off-target cannula (misses) in these three birds were substantially more rostral (1,040 ± 65 mm) than cannula categorized as hits (Supplemental Table S1).
For statistical analyses, birds were categorized into their experimental groups based on whether NE was administered in the NCM. The bird classified as having a unilateral hit was administered PBS and categorized as part of the control group. The two birds considered as bilateral misses were both administered NE during song tutoring and were either considered part of the control group (i.e., no NE infused into the NCM) or as a separate group. Analyses considering "misses" as a separate group showed the same direction of effects and significance as analyses in which "misses" were considered as part of the "control" group (Supplemental Fig. S3).

Statistical Analyses
We used generalized linear mixed effects models (GLMMs; fitted by maximum likelihood using the Laplace approximation) to analyze the effect of NE infusions in NCM on vocal learning (glmer function in the lme4 package: Ref. 64). For example, to examine the effect of NE infusions on vocal learning in song-naïve juveniles, we analyzed the effect of Condition (e.g., NE vs. control) on similarity scores. There was no significant difference in the degree to which songnaïve juveniles learned the songs of the two tutors (F 1,11 = 0.8, P = 0.4013); as such, tutorID was not included in statistical models. However, because multiple pupils could share the same father and mother, NestID was included as a random effect. Similarity scores were normally distributed and, thus, fitted with a Gaussian distribution and the identity linking function. The significance of main effects was assessed using Wald Z test. The same model was used to analyze the percent of pupil syllables shared with the tutor and the number of syllables in the pupil's song. Because there was some variation in the syllable composition of the first motif of the pupils' songs, the number of pupil syllables in the motif was computed as the average syllable counts across the 30 renditions analyzed. To assess the significance of learning, we subtracted similarity scores for tutored birds by the average similarity of the songs of untutored pupils to each tutor (see Song Recording and Analysis above), and then used t tests to determine whether the distributions of differences were significantly different from 0 (see Supplemental Fig. S4 for a complementary analysis in which untutored birds are individually represented). Comparable mixed effects models and t tests were used to analyze the effect of NE on song learning in song-experienced birds (i.e., raised hearing father's song; see also Supplemental Fig. S4).
All statistics were performed using RStudio with a = 0.05.

Effects of NE in NCM on Song Learning in Song-Naïve Juveniles
We first analyzed the degree to which pairing passive song playback with NE infusions into the NCM affected vocal learning in juveniles that were previously untutored ("songnaïve juveniles"; see METHODS). Spectrograms of the adult songs of a pupil with NE infusions into the NCM during tutoring ("NE pupil"), of a control pupil, and of the tutor are plotted in Fig. 2A. For this example, the songs of the NE and control pupil were scored to be 65% and 32% similar, respectively, to the tutor's song. This pattern was consistent across tutored birds. Consequently, compared to control pupils (n = 6), NE pupils (n = 7) produced songs that were significantly more similar to the tutor song (mean ± SE: NE: 54.6 ± 6.5%; control: 26.0 ± 6.6%; v 2 1 = 14.5, P = 0.0001). [Significant differences were also obtained in mixed effects models when the two animals that were considered "misses" were removed from the control group (v 2 1 = 10.4, P = 0.0012) and when these two animals were considered in a separate category of misses (v 2 1 = 12.67, P = 0.0017; Supplemental Fig. S3).] We also assessed the significance of learning within NE and control birds (see METHODS) and found that the extent of song learning was significant for NE birds but not for control birds (t test; H 0 : mean = 0; NE: t 6 = 5.7, P = 0.0013; control: t 5 = 1.3, P = 0.2479; Fig. 2B). Identical patterns of significant learning were obtained when untutored birds were considered individually in the analysis (see Supplemental Fig. S4).
Relatedly, NE pupils produced songs as adults that contained a greater number of syllables copied from their tutors (v 2 1 = 21.8, P < 0.0001; Fig. 2C). [Significant differences remained when the two animals that were considered "misses" were removed from the control group (v 2 1 = 15.4, P < 0.0001) or were considered as a separate group: (v 2 1 = 22.2, P < 0.0001); Supplemental Fig. S3]. On average, NE pupils imitated 3.4 ± 0.5 tutor syllables in their songs whereas control pupils produced 1.2 ± 0.6 tutor syllables. The motifs of NE birds also contained significantly more syllables than those of control birds (v 2 1 = 12.7, P = 0.0004). Importantly, the difference in tutor syllables learned remained significant even after normalizing for the number of syllables in the pupil motif: a greater percentage of NE pupil syllables were considered to be learned from the tutor than control pupil syllables (NE: 66.5 ± 9.5%; Control: 31.6 ± 14.9%; v 2 1 = 7.5, P = 0.0061).  Figure 2. Effects of NE infusions in NCM on song learning in juveniles that were previously naive to song ("song-naïve" juveniles). A: spectrograms of a single motif of one of the two tutor song stimuli that were used to tutor song-naïve juveniles (top), the adult song of a pupil that received NE infusions in NCM during song playbacks (middle), and the adult song of a pupil that did not receive NE in NCM during song playbacks (bottom). Numbers represent an average of automated and subjective measures of song similarity. The adult song of the NE pupil was 65% similar to tutor while the adult song of the control pupil was only 32% similar to tutor. B: birds that received NE in NCM (n = 7) during passive playbacks of tutor song produced adult songs that were significantly more similar to their tutors than the adult songs of pupils that did not receive NE in NCM during tutoring (n = 6). Moreover, comparing the degree of learning of tutor songs against untutored pupils (represented by the black dashed line) revealed that only NE pupils showed significant learning. C: in accordance with their songs being more similar to their tutor's, NE pupils also learned significantly more tutor syllables than control pupils. Squares denote "miss pupils" (see METHODS). Ã P < 0.05; bars and error bars are mean þ SE. NCM, caudomedial nidopallium; NE, norepinephrine.
It is possible that, by chance alone, NE pupils produced songs that were more similar to tutor's song even before tutoring than control pupils. To explore this possibility, we analyzed the similarity of songs produced by developing juveniles on the day before the first day of tutoring. Because motifs cannot be identified in the songs of young juveniles, we spliced out the first second of 30 randomly selected renditions of juvenile song and analyzed the similarity of these segments of song to the tutor motif using SAP. SAP similarity scores of songs before tutoring were not significantly different between NE and control pupils (NE: 36.2 ± 4.6%; Control: 35.9 ± 3.4%; v 2 1 = 0.0, P = 0.9551).

Effects of NE in NCM on Song Learning in Song-Experienced Juveniles
Sequential tutoring has been used as a method to investigate the timing and extent of developmental song learning (reviewed in Ref. 49). Raising birds with their father ("songexperienced juveniles") but socially tutoring them with a tutor song that is acoustically distinct from their father's song can lead to a range of outcomes. For example, sequential tutoring can lead to a pupil producing a song that matches the tutor's song and bears little resemblance to the father's song. In other cases, sequential tutoring can lead to the "hybridization" of the father and tutor's songs, in which case a pupil's song contains song elements from both the father and tutor's song. Regardless of the nature of song learning, previous studies have observed that prolonged periods (e.g., several weeks) of social tutoring can lead to the incorporation of tutor syllables into the adult songs of song-experienced pupils.
Given that NE infusions in NCM potentiated learning in song-naïve juveniles, we hypothesized that NE infusions may similarly affect the learning of a novel tutor's song in song-experienced juveniles. To this end, we analyzed the extent to which NE infusions into the NCM could enhance song learning of a second tutor in song-experienced juveniles. Using the same paradigm used for tutoring song-naïve juveniles, we tutored song-experienced juveniles (n = 8 birds) for 5 days with passive playbacks of song stimuli (n = 30 songs) from a tutor that produced a song that was acoustically distinct from their father's song (tutor 1 from studies with song-naïve juveniles, see METHODS). As adults, songexperienced pupils produced songs that generally resembled their father's song. An example is provided in Fig. 3A. These two pupils were raised by the same father, but one was administered NE into the NCM during song tutoring whereas the other received infusions of PBS ("control"). The adult songs of these NE and control birds were rated to be comparably similar to their father's song (68% and 75%, respectively). Across pupils, the extent of song similarity to their father's song was not statistically different between NE (n = 5) and control pupils (n = 3; v 2 1 = 0.1, P = 0.7218; Fig. 3B). When viewed together, pupils produced songs that were significantly more similar to their father's song than chance (t 7 = 4.6; P = 0.0018). NE and control pupils also incorporated a comparable number of father syllables into their adult songs (v 2 1 = 1.5, P = 0.2232; Fig. 3C). In contrast to the lack of difference in the similarity to their fathers' song, NE and control birds that were sequentially tutored produced songs that differed in their similarity to the tutor's song. In the example provided in Fig. 4A, the adult song of the NE pupil was 36% similar to the tutor song and incorporated two tutor syllables, whereas the adult song of the control pupil was only 19% similar to the tutor song and did not incorporate any tutor syllables. Across all sequentially tutored birds, NE pupils were found to produce songs that were significantly more similar to the tutor song than control pupils (NE: 28.7 ± 2.3%; control: 19.3 ± 0.5%; v 2 1 = 9.6, P = 0.0020; Fig. 4B). Similarly, the extent of song learning from the tutor was significantly greater than chance for NE pupils but not for control pupils (NE: t 4 = 3.8, P = 0.0188; control: t 2 = 1.0, P = 0.4049; see also Supplemental Fig. S4).
Relatedly, NE pupils tended to incorporate more of the tutor's syllables into their songs than control pupils (v 2 1 = 3.2, P = 0.0723; Fig. 4C). Although NE pupils tended to produce more syllables in their motif (NE: 5.  Figure 3. Effect of NE in NCM on song learning in juveniles previously tutored by their fathers ("song-experienced" juveniles). A: spectrograms of a single motif of the song of the father (top), the adult song of a pupil that was raised by the father and that received NE infusions in NCM during song playbacks of another tutor song, and the adult song of a pupil that was raised by the father and did not receive NE in NCM during song playbacks. Numbers represent an average of automated and subjective measures of song similarity. The adult song of the NE pupil was 68% similar to father while the adult song of the control pupil was 75% similar. B: both birds that did (n = 5) and did not receive NE in NCM (n = 3) during passive playbacks of tutor song demonstrated learning (comparison against untutored pupils represented by dashed black line) and learned to produce adult songs that were equally similar to their fathers. C: both groups of pupils incorporated similar numbers of father syllables into their adult songs. Ã P < 0.05, $P < 0.10, bars and error bars are mean ± SE. NCM, caudomedial nidopallium; NE, norepinephrine. v 2 1 = 3.0, P = 0.0841), this trend remained even after controlling for the number of syllables in the motif (i.e., a greater proportion of NE pupil syllables were considered learned from the tutor; NE: 13.9 ± 6.2%; Control: 4.9 ± 4.9%; v 2 1 = 3.0, P = 0.0849).
Similar to the experiment using song-naïve juveniles, we examined the possibility that, compared to control pupils, NE pupils produced songs that were more similar to the tutor song even before tutoring. SAP similarity scores of songs before tutoring were not significantly different between NE and control pupils (NE: 41.5 ± 3.9%; Control: 33.1 ± 4.0%; v 2 1 = 1.4, P = 0.2331).

Effects of NE in NCM on Song Learning in Song-Naïve and Song-Experienced Juveniles
Both song-naïve and song-experienced juveniles were experimentally tutored for 5 days with one of two different tutor songs (tutor 1 or tutor 2; 30 song playbacks per day) beginning at comparable ages (45-60 days posthatch). Therefore, to obtain a robust measure of NE effects on song learning, we analyzed song similarity scores of NE and control pupils in both experiments simultaneously. Specifically, we used a full-factorial model with condition (NE vs. control) and experiment (song-naïve vs. song-experienced) and their interaction as independent variables, and song similarity as the dependent variable (with NestID as a random variable). Overall, NE pupils produced adult songs that were significantly more similar to the tutor's song than control pupils (v 2 1 = 19.1, P < 0.0001), highlighting similar effects of NE across experiments. However, there was a trend for songexperienced pupils to learn less from their tutor's song (v 2 1 = 3.0, P = 0.0845) and a significant interaction between condition and experiment (v 2 1 = 4.0, P = 0.0452). The latter was likely due to the degree of NE-dependent enhancement of learning being larger in song-naïve pupils than in songexperienced pupils. Integrating across experiments, NE pupils learned significantly more tutor syllables than control pupils (v 2 1 = 24.6, P < 0.0001), and a greater proportion of their syllables were considered learned from the tutor (v 2 1 = 9.5, P = 0.0021). Additionally, across experiments, the songs of juvenile NE and control pupils before tutoring did not differ in their similarity to the tutor songs (v 2 1 = 0.3, P = 0.5976).

DISCUSSION
Sensory learning serves as the foundation for the sensorimotor learning of many acquired behaviors. For example, the sensorimotor learning of birdsong during development involves comparing one's current vocal performance to a sensory representation of a "target" song; as such, the development of an accurate imitation of a tutor's song requires learning an accurate sensory representation of the tutor's song (14,16,20,22). Because studies in mammals highlight the importance of NE acting within sensory processing areas to sensory learning and plasticity (37,(42)(43)(44)65), we investigated how infusions of NE into the NCM, an auditory processing area implicated in song learning (16,23,66), affects the fidelity of song learning and imitation. We report that coupling NE infusions into NCM with limited song tutoring increases the accuracy of song learning: juvenile zebra finches that received NE infusions in NCM produced adult songs that were significantly more similar to the tutor song than did juveniles that received control infusions during song tutoring. These novel findings in developing songbirds underscore that NE can potentiate the formation of sensory representations that are used to shape behavioral development.
Pairing NE infusions into NCM with tutoring enhanced vocal learning in juvenile zebra finches that were either song-naïve (i.e., raised with no exposure to song during the critical period for song learning; Fig. 2) or song-experienced (i.e., raised with exposure to song during the critical period for song learning; Fig. 4). However, when song-naïve and song-experienced pupils were analyzed together, both the overall learning and the magnitude of NE enhancement of learning of the experimental tutor's song were reduced for song-experienced pupils compared to song-naïve pupils.
That song-experienced pupils demonstrated less learning from passive playbacks of an experimental tutor's song than song-naïve pupils is consistent with previous findings (reviewed in Ref. 49). For example, Eales (67) reported that song-experienced juveniles learn very little from the songs of a novel tutor that they were passively exposed to for weeks, whereas other studies have noted that song-naïve juveniles will learn from brief epochs of passive tutoring (e.g., Refs. 46 and 68). These results collectively suggest that early learning reduces the ability for subsequently learning from passive song exposure (a relatively weak form of tutoring). In this respect, it is possible that higher concentrations of NE or more prolonged pairing of NE infusions with song tutoring could have led to an enhancement of learning in song-experienced juveniles that was comparable to those observed in song-naïve juveniles. It should be noted that song-naïve juveniles were tutored with the songs of one of two different tutors, whereas song-experienced juveniles were tutored with only the songs of one of these tutors (see METHODS); however, because song-naïve juveniles demonstrated similar learning in response to either tutor's song, we propose that differences in NE effects between song-naïve and song-experienced juveniles were due to difference in prior song experience and not the range of tutor stimuli. These data support the notion that catecholaminergic modulation of activity and plasticity within auditory cortical circuitry are important for the sensory learning of song (29,35,46). These results are consistent with previous findings that NE neurons in the LC are more active during tutoring paradigms that enhance song learning in juvenile zebra finches (46), and although it is important to reveal the effects of NE receptor antagonism on learning, these data suggest that differential NE release into the NCM could contribute to the social enhancement of vocal learning. Indeed, the degree of learning in pupils that received NE infusions (Fig. 2) is similar to that of pupils that were socially tutored for the same number of days (46). These data in juvenile songbirds complement findings in adult songbirds demonstrating that NE modulates auditory responses and the expression of cellular markers of auditory learning in NCM (30)(31)(32)47). Further, because sensorimotor learning occurs over weeks to months and depends on persistent sensory representations of tutor song and because zebra finches produce the song they learn during development throughout their life, these results extend the work in adult songbirds by highlighting how noradrenergic modulation of sensory learning leads to lasting changes in vocal performance. In this respect, these experiments also complement the documented role of NE in other forms of enduring sensory learning, including olfactory imprinting (69,70) and visual imprinting (71).
The precise mechanisms by which NE in NCM enhances the sensory learning of song during development remains unknown, but we hypothesize that changes to the precision or signal-to-noise of neural responses could represent key mechanisms of NE action (72)(73)(74). Norepinephrine increases the reliability and strength of sensory responses in a wide range of vertebrates, including auditory responses in adult songbirds (31,32,(74)(75)(76). Variation in either the precision or signal-to-noise of sensory responses can influence the degree to which sensory experiences drive synaptic plasticity (77)(78)(79); for example, more precise auditory responses can promote spike-timing-dependent plasticity (reviewed in Ref. 77). Although the extent to which NE in NCM modulates auditory responses in juvenile songbirds remains unknown, we propose that NE could enhance the precision of auditory responses and spike-timing-dependent plasticity of NCM neurons in response to song tutoring (see Ref. 34).
Norepinephrine infusions into NCM influence the development of a behavior that depends on lasting representations of the tutor's song, but the precise loci of long-term tutor song storage remain unknown. One possibility is that NE in NCM leads to a more robust and lasting representation of the tutor song in NCM. This model is based on the findings that signaling cascades in NCM are important for the sensory learning of song, that a population of neurons in NCM becomes tuned to the tutor's song after tutoring, and that lesions of NCM eliminate preferences for the tutor song (reviewed in Ref. 14). However, other experiments argue against the notion that the long-term memory of the tutor (target) song resides in NCM and for the importance of sensorimotor circuitry for sensory learning (reviewed in Ref. 22). For example, one study reported that lesions within NCM did not significantly interfere with the accurate sensorimotor development of song (22,80). In addition, a number of experiments have suggested that tutoring-dependent plasticity within the sensorimotor nucleus HVC (used as a proper name) is critical for the sensory learning of song (81)(82)(83)(84)(85). In this respect, it is possible that NE infusions into NCM allow for greater tutoring-dependent plasticity within sensorimotor circuitry. Supporting this notion is the finding that neuromodulation within NCM can affect sensory responses within sensorimotor circuitry (86,87). Revealing how NE manipulations in the NCM of juvenile birds affect auditory responses in NCM and HVC or how song learning is affected by NE infusions in NCM coupled with perturbations of HVC activity will be important to differentiate these possibilities.
Regardless of the precise mechanism of NE action, the current study provides a complementary perspective on the role of catecholamines in the sensory learning of birdsong. A previous study demonstrated that dopamine acting within HVC modulates the extent and fidelity of song learning (85).
In particular, the authors demonstrated that pairing stimulation of dopaminergic terminals in HVC and blocking dopamine signaling in HVC enhanced and attenuated song learning in juvenile zebra finches, respectively. In addition, they revealed that social interactions with an adult tutor that enhance song learning led to greater dopamine release in HVC. In conjunction with previous findings that social tutoring increases the activity of NE neurons in the LC and our current findings, the data collectively suggest that multiple types of catecholamines acting within distinct brain regions (NE acting in NCM or dopamine acting in HVC) could mediate the enhancement of song learning through social interactions with an adult tutor.
Taken together, our studies document, for the first time, that NE manipulations in sensory circuitry affect vocal development in songbirds. It will be important for future studies to extend our findings by examining how antagonism of NE receptors influences the fidelity of song learning and revealing the receptor subtypes that mediated the observed effect. In addition, it will also be important to examine NE effects on neurophysiological responses and representations in developing songbirds as well as the temporal dynamics of NE contributions to sensory learning of song.