HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 92/5/2771    most recent 00467.2004v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Yazaki-Sugiyama, Y. Articles by Mooney, R. Search for Related Content PubMed Articles by Yazaki-Sugiyama, Y. Articles by Mooney, R.

Sequential Learning From Multiple Tutors and Serial Retuning of Auditory Neurons in a Brain Area Important to Birdsong Learning

Department of Neurobiology, Duke University School of Medicine, Durham, North Carolina 27710

Submitted 5 May 2004; accepted in final form 22 June 2004

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Songbirds hear many vocal models during a juvenile sensitive period, transiently imitating some while retaining imitations of others in their repertoires. Despite subsequent conflicting experiences, early experience can exert lasting effects on neural structure and function, raising the possibility that transiently expressed vocalizations or their relevant models are stored in the adult songbird's brain. One site where learned song representations could be stored is the lateral magnocellular nucleus of the anterior nidopallium (LMAN), which in the adult songbird contains neurons responsive to playback of the bird's own song (BOS) and the tutor song (TS). To test whether LMAN neurons develop and retain responses to transiently learned songs, we exposed zebra finch hatchlings [posthatch day 0 (PHD0)] to a first TS (TS1) for about 30 days, isolated them for about 30 days, then exposed them to a second TS (TS2) for 30 days starting at PHD 60. Behavioral analysis showed that PHD 60 juveniles had started to copy TS1, although this copying was transient, because the adult BOS resembled TS2 and not TS1. We found that LMAN auditory responses paralleled these behavioral changes: LMAN neurons at PHD 60 responded strongly and selectively to both the juvenile BOS and TS1, whereas LMAN neurons in adults responded to the adult BOS and TS2, but not to the transiently learned song or its model. Therefore LMAN auditory responses can be lost or overwritten as the juvenile copies a new song, suggesting that the adult LMAN does not store information about transiently learned songs or their models.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Conflicting experiences in the juvenile can result in multiple sensory representations and enhanced behavioral capabilities in the adult (Feldman and Knudsen 1997; Knudsen 1998; Tees and Werker 1984; Udin 1983). Unlike most animals, songbirds learn their vocal repertoires by imitating vocalizations heard during a juvenile sensitive period (Doupe and Kuhl 1999; Marler and Peters 1987). Of many models heard, some are imitated only transiently whereas others persist in the adult repertoire (Marler and Peters 1981). This study addresses whether songs transiently learned by juvenile zebra finches evoke auditory responses in a nucleus important to song learning.

The telencephalic nucleus lateral magnocellular nucleus of the anterior nidopallium (LMAN) is necessary for juvenile song learning (Bottjer et al. 1984; Scharff and Nottebohm 1991) and adult forms of audition-dependent vocal plasticity (Brainard and Doupe 2000) (Fig. 1). LMAN neurons exhibit both auditory and vocal-related activity (Doupe and Konishi 1991; Hessler and Doupe 1999), and juvenile auditory and vocal experience influences the structure and function of LMAN neurons. Tutor song experience acts on or through LMAN because raising birds in isolation from a tutor slows LMAN synaptic maturation (Livingston et al. 2000; Wallhausser-Franke et al. 1995) and blocking N-methyl-D-aspartate (NMDA) receptors in LMAN during tutoring impairs copying (Basham et al. 1996). Auditory and vocal experience also strongly influences the response properties of LMAN neurons, which are highly selective for playback of the bird's own song (BOS) and tutor song (TS) versus other conspecific songs (CON) (Doupe and Konishi 1991; Solis and Doupe 1999). Auditory experience of TS and auditory–vocal experience of BOS shapes these responses because juveniles that sing disrupted songs as a result of vocal nerve section contain LMAN neurons responsive to both TS and BOS (Solis and Doupe 1999). Notably, after these birds mature, BOS and TS playback largely fail to evoke suprathreshold responses from LMAN neurons, even though these stimuli activate areas presynaptic to LMAN (i.e., area X) (Solis and Doupe 2000). In other systems, synapses patterned by early experience are functionally suppressed when in conflict with subsequent experience (Brickley et al. 1994; Zheng and Knudsen 1999), raising the possibility that LMAN auditory responses remain subthreshold when the BOS conflicts with the stored model.

View larger version (24K): [in this window] [in a new window]   FIG. 1. Zebra finch song development, sequential tutoring, and the lateral magnocellular nucleus of the anterior nidopallium (LMAN) recording methods used in this study. A: in normal zebra finch song development, males memorize a tutor song (TS) model during a sensory acquisition phase between posthatch days (PHD) 20 and 60, and vocally copy this model during a sensorimotor learning phase, PHD 40–90. Around PHD 90 the plastic song of the juvenile zebra finch becomes acoustically stable, a process referred to as "crystallization." In the sequential tutoring protocol used here, juvenile male zebra finches were exposed to the song of the first tutor (TS1) until PHD 30 and then isolated for 30 days, then exposed to the song of a second tutor (TS2) for 30 days, followed by another isolation period. Songs were recorded just before and during the exposure to the second tutor, and also after the end of the second isolation period (>PHD 115). B: intracellular recordings were made from LMAN neurons and free-field song playback was used to measure their auditory responses. LMAN neurons from urethane-anesthetized adult birds (>115 PHD) were tested with various song stimuli, including the bird's own song (BOSAdult), the BOS produced at PHD 60 (just before introduction of the second tutor; BOSPHD60), the 2 tutor songs (TS1 and TS2), these various songs played in reverse (Song-REV), and other zebra finch songs [conspecific song (CON)]. Recordings from LMAN neurons in another set of juvenile birds were made at the end of the first isolation period (PHD 60). Song stimuli used for these birds included BOSPHD60, TS1, REV, and CON. Current-clamp records (shown as trace to the right of the recording electrode) were used to assess subthreshold and action potential responses. Connectivity of LMAN is shown in the flowchart within the balloon; HVC (used here as a proper name), RA (robust nucleus of the arcopallium), and the VRN (brain stem vocal respiratory network) are obligatory for singing and form a descending pathway that controls muscles of the avian organ of song, known as the syrinx. LMAN is part of an anterior forebrain pathway that is essential to audition-dependent vocal plasticity, and receives indirect auditory input from HVC, through intervening structures including a basal ganglia homologue area X and the medial dorsolateral (DLM) thalamic nucleus.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Behavioral and electrophysiological techniques have been extensively described in previously published studies (Livingston et al. 2000; Rosen and Mooney 2000) and so are described more briefly here. All experiments were performed in accordance with a protocol approved by the Duke University Instructional Animal Care and Use Committee.

Sequential tutoring protocol

Thirty-four male zebra finches (Taeniopygia guttata) were reared using a protocol that carefully controls auditory and social exposure to adult tutors, and that has been shown previously to successfully extend sensory acquisition beyond posthatch day (PHD) 60 (Livingston et al. 2000). Twenty-three of these experimental birds were hatched and reared until PHD 22–36 (28 ± 0.82 PHD, n = 23) with their father (zebra finch–fathered birds: ZF-birds), mother, and siblings in a sound isolation chamber (Industrial Acoustic, Bronx, NY). Eleven of the 34 experimental birds were reared initially by Bengalese finch (Lonchura striata) foster "parents" until PHD 25–35 (Bengalese finch–fathered birds: BF-birds; 31 ± 1.3 days; n = 11 BF birds). The Bengalese finch song differs from zebra finch song in numerous ways, including shorter duration of individual syllables, a tendency to produce notes with broadly spaced harmonic elements, a higher incidence of single-syllable repeats, and a variable syllable sequence. Therefore by using a Bengalese finch as a foster father we could enlarge the acoustical differences between the first and second tutor songs (TS1 and TS2), making it easier to assess which tutor supplied material copied in the adult song of the sequentially tutored bird. Beginning PHD 22 to 36, individual birds were placed in a small cage, with 2 to 6 of these cages housed in a sound isolation chamber. During this period, BF birds and ZF birds were never housed in the same isolation chamber. Opaque partitions were placed between individual cages. Juveniles housed in this manner could hear but not see each other, and were totally isolated from adult birds of either sex. Under these conditions, juvenile zebra finches do not copy from each other (Livingston et al. 2000; Morrison and Nottebohm 1993). Although birds were isolated from the first tutor across a fairly wide age range (i.e., PHD 22–36), no correlation was detected between the age of isolation from the first tutor and the acoustic similarity of the BOS at PHD60 and TS1 using either Sound Analysis or subjective scoring methods (data not shown; see following text for song analysis methods). Birds remained in visual isolation until PHD 60–65, when they were either used for experiments or subjected to a second tutor. Twelve birds were either used for electrophysiology at PHD 60 (7 ZF and 1 BF), or otherwise were not subjected to a second tutor, and used strictly for behavioral analysis (4 BF birds). The remaining 22 birds were paired with a second tutor (an adult male zebra finch distinct from the biological or foster father) starting at PHD 60 (mean = 61.3 ± 0.42) for 20–34 days (mean 24.1 ± 0.94 days), and then isolated again until used for electrophysiological recordings after PHD 115 (mean = 139 ± 4.8) (Fig. 1A). Adult birds that had been sequentially tutored by 2 zebra finches are termed "ZZ" birds (n = 16), whereas adult birds initially tutored by a Bengalese finch and then by a zebra finch are termed "BZ" (n = 6) birds. See Table 1 for the behavioral analysis of these birds.

Songs were recorded at approximately 1-wk intervals beginning just before introduction of the second tutor (PHD 60) for 1 mo, and then approximately biweekly continuing up to the day before electrophysiological recording. This recording schedule was followed because we observed that the juvenile bird's songs changed dramatically in the first 2 wk of the second tutoring period, and then changed more gradually toward the end of this tutoring period and throughout the second isolation period. To obtain song, the bird was placed in a small recording chamber with an adult female except during the second tutoring period, when the bird was recorded in the presence of its tutor. We observed that the time juveniles started to sing after introduction to the recording chamber could be quite variable, from several hours to more than a day. Therefore we recorded the bird's song during the second tutoring period in the presence of the second tutor, to minimize interindividual variation in the amount of total tutor exposure. All songs were recorded using a Shure SM-57 microphone placed about 25 cm in front of the bird's cage, the signal from which was low-pass filtered at 10 kHz, sampled at 22.05 kHz using a data-acquisition card (National Instruments AT MIO 16E Series), and written directly to disk using software written in the LabView environment by R. Balu. The songs from the tutor birds, as well as other unrelated conspecific songs (CON), were recorded in the presence of an adult female. Before electrophysiological experiments in adult birds, songs were recorded at least twice over an interval of one or more weeks to ensure that birds were singing stable songs. Final songs were recorded within 24 h preceding the experiment.

We used "Sound Analysis" (Tchernichovski et al. 2000) to measure the overall similarity of the BOS at the final adult point (BOS_Adult) (i.e., at the time of electrophysiological recordings we made in adult birds, range 117–205 PHD; mean = 139 ± 4.8; n = 22 birds subjected to electrophysiological recording) to juvenile versions of the BOS [i.e., BOS at PHD 60 (BOS_PHD60)] or to either of the tutor songs (TS1 and TS2), as well as CON songs. When comparing the song of a subject zebra finch to another zebra finch song, we recorded more than 10 renditions of each song and picked the most frequently repeated single motif. We compared single motifs from each of the 2 songs (excluding introductory notes). The various parts of a given motif were identified automatically by periods of silence >70 ms, and then were compared with the parts of the other motif (Sound Analysis settings: "chunk-to-chunk" mode, threshold 81%, interval 70 ms). In Sound Analysis, song similarity is automatically calculated by estimating and comparing 4 acoustic parameters in the 2 songs: pitch, entropy, frequency modulation (FM), and continuity. Here we report the "similarity" scores based on these 4 parameters. For comparisons of juvenile zebra finch songs in which the first tutor was a Bengalese finch, we estimated the average similarity between a BOS_PHD60 and each of 3 renditions of the Bengalese tutor song, because the motif was not clearly defined in Bengalese finch song. In addition, because Bengalese finch songs constitute more tonal syllables of short duration compared with the zebra finch, we used different settings for the interval (39 ms), although the same threshold setting (81%).

Sound Analysis is insufficient by itself to assess low levels of copying, given that 2 songs could achieve a low similarity score either because they were unrelated (i.e., not derived from the same tutor) or because one was modeled only in part after the other. Therefore to better assess copying, we used subjective scoring by 4 observers experienced in birdsong analysis who were asked to compare each of the 2 tutor songs with BOS_PHD60 and BOS_Adult. The observers were blind to the order of the 2 tutors (i.e., which was TS1 or TS2). Observers were instructed to inspect sonograms and listen to each of the 4 songs as needed, then score those syllables of the BOS_Adult or BOS_PHD60, if any, that were similar to those found in either of the tutor songs. The criterion for deeming that a syllable had been "copied" was that at least 3 of the 4 observers scored a syllable of BOS_Adult or BOS_PHD60 as copied from either of the 2 tutors. Percentage similarity of song was calculated from the number of syllables copied from a given tutor divided by the total number of syllables in the BOS_Adult.

Song stimuli for electrophysiological recordings

All songs were low-pass filtered at 10 kHz, digitized at 22.05 kHz, and edited to include 2 or 3 motifs for each song. Song stimuli included the BOS_Adult, BOS_PHD60, TS1, and TS2. Most cells were presented these 4 songs in both the forward and reverse (Song-REV) direction, and also presented a CON song. We used a total of 5 different CON songs for all of the electrophysiological experiments performed here, although we used the same CON song consistently throughout a single experiment. Each stimulus was played through a speaker facing the bird at a distance of about 25 cm and was 1.5–2.5 s in duration, with an average sound pressure level of 70 dB (RMS, A-weighting). The various song stimuli were presented in a fixed sequence during a recording from a given cell.

In vivo electrophysiology

Of the 34 birds used in this project, 22 (16 ZZ and 6 BZ; 1 ZZ bird failed to yield any responsive LMAN neurons) were used for electrophysiological recordings after PHD 115, and 8 (7 ZF and 1 BF) were used for electrophysiological recordings at the end of the first isolation period (PHD 62–69; i.e., they were not exposed to a second tutor). On the day of the electrophysiological experiment, the bird was injected intramuscularly with 20% urethane in dH₂O (90–100 µl total; Sigma, St. Louis, MO) in 30- to 40-µl doses at 30-min intervals, further anesthetized with inhalation anesthetics as needed, and then equipped with a stainless steel post mounted on its head with dental acrylic and cyanoacrylate.

Sharp electrodes (borosilicate glass, BF100-50-10; Sutter Instrument, Navato, CA) were pulled to 60–200 M when filled with 2 M K-acetate, 5% neurobiotin, and placed into the nucleus using stereotaxic coordinates and a hydraulic microdrive (Soma Scientific). An AxoClamp 2B intracellular amplifier (Axon Instruments) recorded intracellular potentials in the bridge mode. The recorded potentials were low-pass filtered at 3 kHz, digitized at either 10 or 11.05 kHz (National Instruments PCI MIO 16E series data-acquisition board) and stored on a PC. In the first several experiments, we used intracellular staining with neurobiotin to confirm post hoc that we were indeed recording from LMAN neurons. In subsequent experiments, LMAN neurons were identified on-line by their characteristic spike shape and firing patterns (Livingston and Mooney 1997; Rosen and Mooney 2000). If the impaled LMAN neuron had a resting membrane potential negative of –60 mV, and also displayed robust spontaneous synaptic activity, we then probed its auditory response properties using the panel of song stimuli mentioned above. In some cases, tonic negative or positive currents were injected through the electrode to shift the membrane potentials of the neurons, to accentuate evoked synaptic or action potential activity, respectively. In these cases, comparisons of responses to different stimuli were made at the same membrane potential. Ten to 90 iterations of each auditory stimulus, including 2 s of silence pre- and poststimulus, were delivered at 5-s intervals to the adults and 10- to 12-s intervals to the young birds by a speaker about 25 cm in front of the bird's head. Typical experiments lasted about 12 h; although robust responses usually persisted over this duration, experiments were discontinued if responsiveness to the BOS diminished or disappeared.

Analysis of electrophysiological data

For analyzing electrophysiological data, we used methods that have been described in previously published studies (Rosen and Mooney 2000). The suprathreshold responsiveness of neurons was calculated by response strength (RS) = S_FR – B_FR, where S_FR and B_FR are the firing rates during each whole song stimulus presentation and during the baseline period before each song stimulus presentation, respectively. To determine whether a suprathreshold response was significant for each cell, we used a paired t-test to compare mean firing rates during each song stimulus presentation and during a prestimulus baseline period of equal duration, and P < 0.05 as the criterion for responsiveness. For measuring the subthreshold response, we calculated a z-score for positive area (Zarea): the area measurements were given by the difference between the average area below median-filtered traces during song stimulus presentation (Sarea) and during the period immediately before each song stimulus presentation (Barea), each calculated relative to the baseline membrane potential value estimated by the mode of the prestimulus resting potential

To maximize the possibility of detecting phasic responses to song playback, we also calculated the Zarea of the response peak: a window (200–800 ms wide) was centered about the highest peak during song presentation, and a window of equivalent size was centered about the largest prestimulus peak. The largest peaks in the median-filtered trace averages were determined by visual inspection. To determine whether a neuron was responsive to a given song at the subthreshold level, we used a paired t-test to compare average area during the peak of each song stimulus presentation and during the peak of a prestimulus baseline period of equal duration, and P < 0.05 as the criterion for significance. To compare the relative response bias to BOS versus one of the other song stimuli [reversed BOS (REV), BOS_PHD60, TS1, TS2, and CON], we used a d' statistic, which estimates the discriminability between 2 stimuli for an ideal observer (Green and Swets 1966)

where RS is the mean value of S_FR – B_FR, and ² is the response variance for the given stimulus. A d' > 0.5 reflects a net bias to the BOS, whereas a d' = 0 reflects no net bias. This measure is superior to a ratio measure of response strengths because it takes into account the variance of the responses and can report negative values (Solis and Doupe 1997).

Statistical tests

In those cases where multiple comparisons were made, an ANOVA was used followed by a post hoc mean comparison using Tukey's test, with significance determined at P < 0.05. Other specific tests used to determine significance are given in the RESULTS and/or figure legends. All values reported are means ± SE, unless otherwise noted.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The first part of this section addresses the behavioral analysis of song learning in juvenile and adult birds raised in the sequential tutoring protocol. The second part of this section describes the electrophysiological response properties of LMAN neurons to song playback in these juvenile and adult birds.

Behavior

SONGS OF PHD 60 JUVENILES RAISED IN A SEQUENTIAL TUTORING PROTOCOL. To distinguish whether copying from TS1 had occurred at the end of the first isolation period (PHD 60), we analyzed song of 15 birds where another zebra finch was the tutor during the first posthatch month ("ZF" birds; note that the song of one ZF bird was not recorded at PHD 60, so this bird was used only for the adult song comparisons). Although song structure is only roughly formed at this stage, complicating subjective analysis, blind observers still identified song syllables similar to syllables in the songs of the first tutor in 3 of the 15 juveniles (Fig. 2; Table 1; mean % similarity = 4.1 ± 2.2%, n = 15 birds; copied syllable range 0–1, mean 0.2 ± 0.11; total syllable range 2–9 syllables, mean 4.5 ± 0.42 syllables). Incontrast, blind observers never scored songs of PHD 60 birds as similar to the second tutor, a bird to which the PHD 60 juveniles had not yet been exposed (Fig. 2C; note that for subjective scoring at this stage the "CON" songs were TS2). A further indication of copying from the first tutor was provided by percentage similarity scores calculated using Sound Analysis, an automated routine that estimates similarity between 2 motifs using 4 acoustic parameters: entropy, pitch, FM, and continuity (Tchernichovski et al. 2000). We found the similarity between BOS_PHD60 and TS1 was significantly higher than that between BOS_PHD60 and TS2 (i.e., a zebra finch song to which they had not yet been exposed) (BOS_PHD60 and TS1 50 ± 6.9%, BOS_PHD60 and TS2, 35 ± 6.3%, P < 0.02, paired t-test; n = 15). In addition, the mean similarity between the BOS_PHD60 and TS1 was slightly but not significantly greater than the mean similarity between the BOS_PHD60 and any of 4 different CON songs (Fig. 2C; P < 0.05, ANOVA). Thus analysis of song before introduction of the second tutor provides evidence of partial copying from the first tutor, at least in some birds.

View larger version (51K): [in this window] [in a new window]   FIG. 2. Songs recorded immediately before introduction of the second tutor showed hints of copying from the first tutor. A: sonograms, generated using Sound Analysis software, depict BOSPHD60 (recorded immediately before exposure to the second tutor) from 2 different juveniles (Bird 1 and Bird 2) and their first tutor's song (TS1; another zebra finch). Some hints of copying from TS1 can be seen in the songs of both juveniles. Syllables in the given BOS that are copied from the tutor song are denoted by the same letter used for the given tutor syllable, and marked by a prime symbol (d' in Bird 2 is copied from d in TS1). The song of Bird 1 showed 50.2 and 25% similarity to TS1 by Sound Analysis and subjective scoring, respectively, whereas the song of Bird 2 showed 58.3 and 17% similarity to TS1 using these 2 methods of analysis. B: song similarity between BOSPHD60 and TS1, as measured using Sound Analysis and by subjective scoring, in all 15 birds in which first tutor was another zebra finch. With subjective scoring, 3 of 15 birds were deemed to have some similarity to TS1, whereas no similarity was detected between BOSPHD60 and the second tutor's song (unrelated birds at PHD 60; data not shown). Numbers in parentheses denote number of songs with identical scores. C: mean similarity scores comparing BOSPHD60 to TS1 or CON using either Sound Analysis or subjective scoring, for the same birds as in B. Sound Analysis comparison used 4 different CON songs, whereas the subjective scoring comparison used a single CON song (TS2, to which the bird had not been exposed). Mean similarity of BOSPHD60 to TS1 was not significantly higher than that to any of 4 CON songs, based on an ANOVA.

View larger version (41K): [in this window] [in a new window]   FIG. 3. Sequential tutoring with Bengalese and then zebra finches revealed sequential copying from the first and second tutor. A: sonograms of a Bengalese finch TS1 (top), the song of the juvenile reared with this tutor (BOSPHD60; second from top), the TS2 that this juvenile was exposed to starting at PHD 60 (i.e., another zebra finch; third from top), and the final song of this sequentially tutored "BZ" bird (BOSAdult; bottom). Bengalese finch song is distinguished from zebra finch song by relatively shorter-duration syllables that are often repeated, features mimicked in the BOSPHD60. After being tutored by a zebra finch (TS2), the song lost these qualities and became zebra finch–like, characterized by long syllables and a defined motif. Copied syllables from either tutor are denoted in the BOS by the same letter used for the given tutor syllable, and marked by a prime symbol. This bird's song showed 71.8% (Sound Analysis) and 60% (subjective scoring) similarity to TS1 and 33.4 and 0% to TS2 at 60 days. At 120 days, his song (BOSAdult) showed 25.1 and 0% similarity to TS1 and 75.1 and 86% similarity to TS2, using Sound Analysis or subjective scoring methods. B: individual similarity scores of BOSPHD60 and TS1 measured using Sound Analysis and subjective scoring (numbers in parentheses in left graph denote number of songs with identical scores). C: mean similarity scores using these 2 measures for BOSPHD60 and TS1 or CON (right), for juvenile zebra finches initially tutored by a Bengalese finch (n = 11). In the Sound Analysis comparison, BOSPHD60 was more similar to TS1 than to CON songs, as determined by an ANOVA. In the subjective scoring comparison, the BOSPHD60 was more similar to TS1 than to CON (i.e., TS2; paired t-test, P < 0.01).

ADULT SONGS OF SEQUENTIALLY TUTORED BIRDS. Adult birds that had been sequentially tutored as juveniles by 2 birds showed strong evidence of copying from the most recent tutor (TS2), and little or no sign of copying from TS1 (Figs. 3A and 4A). We exposed a total of 22 zebra finches to a second (zebra finch) tutor starting at 60 days [mean 61.3 ± 0.42 PHD; 16 birds were initially tutored by another zebra finch (ZZ birds), and another 6 birds were initially tutored by a Bengalese finch (BZ birds)]. Syllable copying from the second tutor, as assessed by subjective scoring, was observed in 14/16 adult ZZ and 4/6 adult BZ birds (Fig. 4B). For all 22 birds, we found that on average 2.9 ± 0.4 out of 6.2 ± 0.4 BOS syllables were copied from TS2 (Table 1), yielding a mean subjective similarity score of 48.7 ± 7.0% (Fig. 4). Automated scoring of similarity between BOS_Adult and TS2, using Sound Analysis, showed that the BOS_Adult was significantly more similar to TS2 than to the songs of nontutor (CON) zebra finches (61.5 ± 5.6% vs. 36.3 ± 2.2%; each BOS_Adult was compared with the songs of 4 nontutor zebra finches from our colony). The similarity scores between BOS_Adult to TS2 obtained using Sound Analysis (>60% similarity) were comparable to the similarity scores others have obtained when comparing an adult bird's song to that of its tutor (Tchernichovski et al. 2001).

View larger version (43K): [in this window] [in a new window]   FIG. 4. Copied syllables in the adult song of sequentially tutored birds were from the most recent tutor (TS2), and not from TS1. A: examples of songs from a sequentially tutored zebra finch and his 2 zebra finch tutors. Sonograms depict the BOSPHD60, the BOSAdult, the TS1, and the TS2. Syllables in the BOSAdult that were copied from TS2, as determined by subjective scoring, were labeled with the same letter marked by a prime symbol. In this example, the majority of subjective scorers detected no similarity between the BOSPHD60 and TS1. BOSAdult showed 44.9 and 0% similarity to TS1 based on Sound Analysis and subjective scoring, respectively, and 62 and 87% (6 of 7 syllables copied) similarity to TS2 by Sound Analysis and subjective scoring. B: song similarity scores between the BOSAdult of 22 sequentially tutored adult birds and their 2 tutors (TS1 and TS2) measured by Sound Analysis (left) or by observers that were blind to the tutoring sequence (right; for details see METHODS). Songs of 15/20 birds were more similar to TS2 than TS1 based on similarity scores using Sound Analysis. Subjective scores indicated that the BOSAdult showed similarity to TS2 in 17 birds, while failing to find similarity to TS1 in 18/22 birds. Filled symbols represent birds that were sequentially tutored by 2 zebra finches ("ZZ" birds); open symbols represent birds that were tutored initially by a Bengalese finch and then by a zebra finch ("BZ" birds). Numbers in parentheses denote number of songs with identical scores.

We found little evidence that adult birds retained syllables from their first tutor after exposure to a second tutor. Over 80% (18/22) of the sequentially tutored adult birds showed no evidence of syllables copied from TS1 (Fig. 4B, right; 4/22 birds retained syllables from TS1: one ZZ bird copied one syllable, one BZ bird copied 2 syllables, and 2 BZ birds copied one syllable; based on subjective scoring). Furthermore, estimates of song similarity calculated using Sound Analysis indicated that the mean similarity between BOS_Adult and TS1 was not significantly different from the similarity between BOS_Adult and CON (48.7 ± 7.0% vs. 36.3 ± 2.2%; NS, ANOVA). However, the incidence of high-similarity scores (i.e., >50% similarity) in the BOS_Adult and TS1 comparisons (8/22 birds) was slightly higher than the range of high-similarity scores in other BOS_Adult and CON comparisons (BOS_Adult >50% similarity to 4 CON songs ranged from 3 to 7 out of the 22 birds, using Sound Analysis). Overall, the BOS_Adult tended to resemble the TS2 and not the TS1, as reflected in the mean similarity scores obtained using both subjective scoring and Sound Analysis (Subjective: BOS_Adult and TS2, 48.7 ± 7.0%; BOS_Adult and TS1, 3.2 ± 1.6%; Sound Analysis: BOS_Adult and TS2, 61.5 ± 5.6%; BOS_Adult and TS1, 42.0 ± 5.2%).

The similarity to TS2 and not TS1 also characterized the songs of the subset of adult birds that previously had shown clear evidence of copying from their first tutor at PHD 60, based on subjective analysis [n = 7; 3 ZZ and 4 BZ birds; 6/7 adult birds showed copying from TS2; average similarity for all 7 birds' songs to TS2 using Sound Analysis was 59.8 ± 11.9%; note that 2 of these birds' adult songs were judged by subjective scoring to be slightly similar to TS1 (25 and 11%)]. These results indicate that the protocol used in these experiments produces adults with songs copied from the most recent tutor (TS2), with little or no copying from the first tutor to whom they were exposed (TS1). This tendency to copy from the most recent tutor is a trait also manifested in zebra finches tutored serially in a contiguous manner [i.e., without an intervening isolation period; see Eales (1985)].

To further quantify how the BOS changed over development in relation to each of the birds' 2 tutors, we estimated the change in similarity between the BOS and a given tutor song from PHD 60 to after PHD 115 (Fig. 5A). These developmental changes in individual birds' songs were estimated by subtracting the similarity score between the BOS_PHD60 and a given tutor song (i.e., TS1 or TS2) from the similarity score between the BOS_Adult and the same tutor song. Automated similarity score differences showed that the songs of 16/21 birds became more similar to TS2 after exposure to the second tutor (Fig. 5A, left). Similarly, subjective similarity score differences revealed that songs of 17/21 sequentially tutored birds became more like TS2 over this period (Fig. 5A, right; note that song from one bird was not recorded before the second tutoring period). In contrast, subjective analysis indicated that similarity to TS1 either stayed the same or decreased in 18/21 birds over this same period (Fig. 5A, right). Comparisons of Sound Analysis similarity score differences also revealed that the majority (13/21) of birds' songs decreased in similarity to TS1 over development (Fig. 5A, left). Ultimately, the mean similarity score of the BOS to TS2 estimated using Sound Analysis increased substantially between PHD 60 and 115 (33.5 ± 4.8% similarity between BOS_PHD60 and TS2 vs. 61.5 ± 5.6% between BOS_Adult and TS2, P = 0.003, paired t-test), whereas over the same period the BOS decreased slightly but not significantly in similarity to TS1 (48.6 ± 6.4% between BOS_PHD60 and TS1 vs. 42.0 ± 5.2% between BOS_Adult and TS1; NS, P = 0.3, paired t-test). Therefore during development, the BOS increased in similarity to TS2, and most often decreased in similarity to TS1. This result suggests that sequentially tutored birds lost or overwrote their old songs as they learned new songs.

View larger version (26K): [in this window] [in a new window]   FIG. 5. BOS changes during and after exposure to the second tutor, becoming more like the second tutor's song. A: difference scores show that the songs of most sequentially tutored birds became more like TS2 between PHD 60 and 115, whereas similarity of most of these songs to TS1 either decreased or stayed the same. Difference scores for each bird were calculated by subtracting the similarity score of the BOSPHD60 to a given tutor song from the similarity score of the BOSAdult and that same tutor song; positive values indicate increased similarity over development. Differences in similarity were calculated for Sound Analysis (left) and subjective scoring (right). ZZ and BZ birds are marked by open circles and filled stars, respectively. Numbers in parentheses denote number of scores. B: timing of the changes to the BOS and the onset of song stabilization in sequentially tutored birds. Sound Analysis was used to estimate the similarity of the BOS recorded before, during, and after the second tutoring period to the BOSAdult. The BOS started to change soon after introduction of the second tutor, and was not significantly different from the BOSAdult within 2 mo after beginning of second tutor period (time points where the similarity of the current BOS is significantly different from the final BOS are marked by asterisks; number of birds recorded at each time is given in parentheses). Approximate extent of second tutoring period is depicted by the cross-hatched box.

Electrophysiological properties of LMAN neurons

JUVENILE LMAN NEURONAL RESPONSIVENESS. Prior studies suggest that LMAN neurons in juvenile zebra finches raised with normal access to a tutor are already partially selective for the bird's own song and the tutor song by PHD 60 (Solis and Doupe 1997). We were curious to know whether LMAN neurons in PHD 60 birds subjected to more restricted tutor access (i.e., for only the first 35 PHD), as in the first phase of the sequential tutoring protocol used here, exhibited selectivity for their own song as well as the TS1. In the event that responses to each song might manifest themselves at only a subthreshold synaptic level, and not in the action potential discharge of LMAN neurons, we used in vivo intracellular recordings and signal-averaging methods to detect both sub- and suprathreshold activity (Fig. 1B; see METHODS).

We recorded LMAN neuronal activity from a set of juvenile birds at the end of their first isolation period (mean = PHD 63 ± 0.98, n = 8 birds, 7 ZF and 1 BF). At PHD 60, we found that BOS_PHD60 and the tutor's song (i.e., TS1) could elicit significant suprathreshold responses in LMAN neurons (Fig. 6A; see METHODS). All 37 LMAN neurons we sampled at this time point responded to the BOS_PHD60, and almost half presented with TS1 responded to this stimulus (12/25 neurons; 12 neurons in 3 birds were not tested with TS1) (Fig. 6A and Table 2). Four of 5 birds tested contained TS1-responsive cells; 3 of these 4 birds copied some material from TS1, whereas the one bird that did not contain any TS1-responsive cells also did not copy from TS1, as assessed by subjective scoring. The mean response strength was greater for the current BOS (i.e., BOS_PHD60) than that for all other stimuli tested (Fig. 6B; see figure for statistical comparisons). We also assessed the neuronal response at a subthreshold level, by calculating positive area z-scores over the duration of the entire song stimulus (Z_area; see METHODS). In general, the subthreshold response properties were the same as those at the suprathreshold level (Fig. 6A). We encountered a total of 35 neurons that showed significant subthreshold responses. All 35 neurons responded to BOS and almost half presented with TS1 responded to this stimulus (12/25 cells; 12 neurons in 3 birds were not tested with TS1). The mean subthreshold response z-score was greatest for the BOS and the rank order of stimulus-evoked subthreshold response strengths was identical to the pattern seen with suprathreshold response strength measurements (Fig. 6C). These results indicate that the current BOS and TS1 are effective at driving sub- and suprathreshold responses in LMAN neurons of PHD 60 zebra finches tutored with this protocol, and that CON, BOS–REV, and TS1–REV are largely ineffective stimuli.

View larger version (28K): [in this window] [in a new window]   FIG. 6. LMAN neurons in juvenile (PHD 60) birds are strongly and selectively responsive to the current BOS (BOSPHD60) and also respond to the tutor song (TS1) heard during the 1st mo posthatch. A: example of song-evoked responses from an LMAN neuron in a PHD 60 juvenile bird that had been tutored with a zebra finch for the first posthatch month, then isolated for another month before the electrophysiological recording. Both the bird's own song (BOSPHD60) and tutor song (TS1) evoked strong subthreshold (averaged membrane potential records; middle) and suprathreshold [cumulative poststimulus time histogram (PSTH); top] responses, whereas another zebra finch song (CON) that the juvenile bird had not been exposed to previously failed to evoke a response. Response strength (RS; spikes/s) and P values for the suprathreshold responses are shown above the PSTH for each stimulus. PSTH bin size is 25 ms; data are from 30 iterations of each song stimulus, shown as an oscillogram at the bottom. B: mean RS values (spikes/s) for the various song stimuli for the population of LMAN neurons in PHD 60 birds that responded to at least one of the test stimuli (n = 37 cells). BOSPHD60 evoked the strongest response in the population of LMAN neurons tested here (asterisk indicates P < 0.05, ANOVA test). C: mean subthreshold responses (Zarea) for various song stimuli for the population of LMAN neurons in PHD 60 birds that responded to at least one of the test stimuli (n = 35 cells). As with the suprathreshold responses, the BOSPHD60 evoked the strongest subthreshold responses in the population of LMAN neurons recorded in PHD 60 birds (asterisk indicates P < 0.05, ANOVA test). D: similarity of the BOSPHD60 to TS1 as measured with subjective scoring did not correlate with the TS1-evoked RS (spikes/s) in PHD 60 birds. This lack of correlation differed from that seen in adult birds with respect to the acoustic similarity and responsiveness to TS2 (see Fig. 10). Responses of neurons recorded from birds exposed to zebra finch tutors are marked by filled squares (ZF), and those from birds initially tutored by a Bengalese finch (BF) are depicted with open circles. E: mean d' values indicate that LMAN neurons in juvenile birds are selective (d' >0.5) for the BOSPHD60 vs. either CON or TS1 (left), and are also selective for forward vs. reverse playback of either the BOSPHD60 or TS1 (right). Mean d' value for BOSPHD60 vs. TS1 was not significantly lower than the d' value for BOSPHD60 vs. CON, as determined by a Student's t-test. F: cumulative distribution of d' values comparing responses evoked by the BOSPHD60 with those evoked by TS1, CON, or BOSPHD60 REV (REV) are largely overlapping, indicating equivalent selectivity for the BOSPHD60 over the other stimuli. Grey box marks non-selective region.

The juvenile responsiveness to TS1 could reflect either broadly tuned properties of relatively nonselective neurons or instead narrowly tuned properties of selective neurons. As noted previously, LMAN neurons at PHD 60 as a population exhibited strong responsiveness to the BOS_PHD60 and TS1, but failed to respond to CON, BOS_PHD60–REV, and TS1–REV, hinting that these neurons were already moderately selective in their response properties (Fig. 6B). We used the d' metric to determine whether LMAN neurons at PHD 60 were selective for forward playback of the current BOS versus other songs (see METHODS; when using the BOS as the referent, d' values > 0.5 reflect a biased response to the BOS over the comparison song). These d' values indicated that LMAN neurons at PHD 60 were strongly selective for the BOS_PHD60 versus CON and BOS_PHD60–REV (Fig. 6, E and F: d' BOS_PHD60–CON = 1.14 ± 0.16, n = 23, d' BOS_PHD60–REV =1.03 ± 0.14, n = 37). We also noted that LMAN neurons in these juvenile birds were selective for BOS_PHD60 versus TS1 (d' BOS_PHD60–TS1 = 0.83 ± 0.13, n = 25). Although the mean d' value for this comparison was lower than BOS_PHD60 versus CON, this difference was not significant (Fig. 6E). Furthermore, the cumulative distributions of d' values comparing BOS_PHD60 to CON, TS1, and REV were not significantly different from one another [Kolmogorov–Smirnov (K-S) test], suggesting that LMAN neurons in juvenile birds were equivalently selective for the current BOS versus these other stimuli (Fig. 6F). However, TS1-responsive neurons were selective for forward over reverse TS1 playback (Fig. 6E; d' TS1–TS1 REV = 0.61 ± 0.15, n = 25). Therefore LMAN neurons in these birds were already highly selective for the current BOS over other song stimuli and showed selectivity for forward over reverse playback of the first tutor song. These results indicate that LMAN neurons in PHD 60 zebra finches reared with limited exposure to a tutor nonetheless exhibit narrowly tuned responses to their own song and the tutor song.

LMAN NEURONAL RESPONSES TO SONG PLAYBACK IN ADULT SEQUENTIALLY TUTORED BIRDS. In other systems, patterns of early sensory experience that conflict with later experience can be maintained in a latent manner at the synaptic level (Feldman and Knudsen 1997; Udin 1983). We tested whether LMAN neurons in sequentially tutored adult birds (n = 22) exhibited any responsiveness to the songs of the first tutor or to the transient copies that the juvenile had made of its first tutor's song. In general, we found that LMAN neurons in sequentially tutored adult birds exhibited sub- and suprathreshold responses to the current BOS_Adult and/or the TS2, but responded less often or not at all to either TS1 or the BOS_PHD60 (Fig. 7; Table 2).

View larger version (30K): [in this window] [in a new window]   FIG. 7. In adult sequentially tutored birds, LMAN neurons responded to playback of the current BOS and the most recent tutor song, but not to earlier versions of the BOS or earlier tutor songs. A representative LMAN neuron from an adult sequentially tutored bird responded to the BOSAdult and TS2, but not to TS1 or the BOSPHD60, at both the subthreshold (average membrane potential records; middle) and suprathreshold levels (cumulative PSTHs; top). RS (spikes/s) and P values, from a paired t-test comparing the stimulus-evoked and baseline firing rates on a trial-by-trial basis, are given above the PSTH for each stimulus condition. PSTH bin size is 25 ms; data are from 21 iterations of song playback; song stimuli are shown as oscillograms at bottom.

View larger version (20K): [in this window] [in a new window]   FIG. 8. Average response strengths of LMAN neurons in adult sequentially tutored birds to juvenile and adult versions of the BOS, as well as to the 2 tutor songs experienced by the bird. A: average RS (spikes/s) from the 84 responsive LMAN neurons we recorded in adult birds, showing that the BOSAdult is the most effective stimulus, followed by TS2, BOSPHD60, CON, and TS1; reverse playback of either the BOSAdult or TS2 (BOSAdult REV and TS2 REV) did not drive responses in these LMAN neurons. At a population level, the BOSAdult elicited significantly stronger responses than any other song, whereas TS2 evoked greater mean responses than all other non-Bos songs (asterisk indicates P < 0.05, ANOVA test). Data are from 21 animals, including 15 ZZ and 6 BZ birds. B: mean RS values in LMAN neurons in those 7 adult birds that had copied from their first tutor (TS1) at PHD 60 (i.e., just before exposure to the second tutor), as determined by subjective scoring (4 of these 7 birds were initially tutored by a Bengalese finch). In this subset of cells (n = 29), the mean suprathreshold response of LMAN neurons was greatest for the BOSAdult and for TS2 (asterisk indicates P < 0.05, ANOVA test). None of these LMAN neurons (0/29 cells) responded to TS1.

We investigated whether "latent" synaptic responses to the TS1 or juvenile versions of the BOS might be maintained at a subthreshold level. The results of this subthreshold analysis closely paralleled the results of the suprathreshold analysis, in that both the BOS_Adult and TS2 evoked a strong response, whereas the TS1 and BOS_PHD60 did not (Fig. 9A; see legend for statistical comparisons). As with the suprathreshold analysis, we did not detect any subthreshold responses to TS1 in adult birds that had copied one or more syllables from TS1 at PHD 60 (Fig. 9B).

View larger version (15K): [in this window] [in a new window]   FIG. 9. Average subthresold responses of LMAN neurons in adult sequentially tutored birds to juvenile and adult versions of the BOS, as well as to the 2 tutor songs experienced by the bird. A: average z-scores for positive area (Z score+area; see METHODS) from the 68 subthreshold responsive LMAN neurons we recorded in adult birds, showing that the BOSAdult is the most effective stimulus, followed by TS2, BOSPHD60, CON, and TS1; BOSAdult–REV and TS2–REV did not drive responses in these LMAN neurons. For CON, REV, and TS2–REV we tested subsets of these 68 neurons: 27, 21, and 21, respectively. At a population level, the BOSAdult elicited significantly stronger responses than any other song, whereas TS2 evoked greater mean responses than those of all other non-Bos songs (asterisk indicates P < 0.05, ANOVA test). Data are from 21 animals, including 15 ZZ and 6 BZ birds. B: mean subthreshold responses (Z score+area) in LMAN neurons in those 7 adult birds that had clearly copied from their first tutor (TS1) at PHD 60 (i.e., just before exposure to the second tutor), as determined by subjective scoring (n = 3 ZZ and 4 BZ birds). In this subset of cells (n = 24), the mean subthreshold responses of LMAN neurons were greatest for the BOSAdult, and greater for TS2 than for BOSPHD60, BOSAdult REV, and TS2 REV (asterisk indicates P < 0.05, ANOVA test). In contrast, none of these LMAN neurons (0/24 cells) responded to TS1. C: subthreshold peak (P) analysis of LMAN neurons in adult sequentially tutored birds detected 2 more TS1-responsive cells, and a much larger number (22) of TS2-responsive cells, as contrasted with a whole motif analysis (W). These results suggest that latent responses to TS1 are not present in the LMAN of adult birds. For peak analysis, we measured the subthreshold response measured during a short time window (200–800 ms; see METHODS), bracketing the peak response evoked by the tutor's song.

CORRELATION BETWEEN SONG SIMILARITY AND RESPONSE STRENGTH. We hypothesized that stronger suprathreshold responses may be evoked in LMAN neurons of adult sequentially tutored birds by those tutor songs or juvenile versions of the BOS acoustically more similar to the BOS_Adult. In this case, the response strength of LMAN neurons to a given song should correlate with the acoustic similarity between the BOS and that song. To investigate this possibility, we plotted the response strength of individual LMAN neurons to TS1, TS2, or BOS_PHD60 as a function of the percentage similarity of the BOS to the given song, as measured with either Sound Analysis or subjective methods (Fig. 10). These plots revealed a significant correlation between acoustic similarities and response strength for TS2 and BOS_PHD60 but not for TS1 (Fig. 10). In the case of TS1, the absence of a correlation could be accounted for by the fact that only a small fraction of LMAN neurons responded to this stimulus. However, we also noted that the acoustic similarities between the BOS_Adult and TS1 estimated using Sound Analysis spanned a range greater than that for the similarities between the BOS_Adult and the BOS_PHD60 (compare top and bottom left, Fig. 10), suggesting that the lack of responses to TS1 was not simply attributed to acoustic dissimilarities between the BOS_Adult and TS1. Therefore acoustic similarity as measured with these 2 techniques can account for LMAN responsiveness for only the most recent tutor song and juvenile versions of the BOS, but not for a tutor song copied only transiently earlier in development.

View larger version (27K): [in this window] [in a new window]   FIG. 10. Response strength (RS) to either the juvenile BOS or the first or second tutor song plotted as a function of the acoustic similarity of that song to the adult BOS. Relationship of RS (spikes/s) to each of the 2 tutor songs that the bird was exposed to during development [TS1 (top) and TS2 (middle)] and old versions of own song (BOSPHD60; bottom), plotted as a function of the similarity of that tutor song to the BOSAdult, calculated using either Sound Analysis (left column) or subjective scoring methods (right column). A significant correlation was detected between the response strength to TS2 or BOSPHD60 and the similarity of either of these songs to the BOSAdult as assessed with either analysis method (middle and bottom). In contrast, no significant correlation was detected between the response strength to TS1 and the acoustic similarity of TS1 to the BOSAdult (top). LMAN neurons from ZZ birds are plotted as filled squares, whereas those from BZ birds are plotted as open circles.

View larger version (19K): [in this window] [in a new window]   FIG. 11. Auditory responses of LMAN neurons in adult sequentially tutored birds are strongly selective for the BOSAdult over other songs. A: cumulative distribution of suprathreshold d' values (see METHODS) for the BOSAdult vs. other songs used as test stimuli, including BOSPHD60, TS1, TS2, CON, and BOSAdult–REV. Distributions were similar for all songs, although the distribution for the BOSAdult vs. TS2 was shifted leftward (in the direction of decreased selectivity; BOSAdult–TS2 vs. BOSAdult–TS1, P < 0.001; BOSAdult–BOSPHD60, P = 0.01; BOSAdult–CON, P = 0.001; Kolmogorov–Smirnov tests). Cells with responses biased to the BOSAdult exhibit d' > 0.5, whereas d' < –0.5 denotes a response bias to the other song. Values between –0.5 and 0.5 (shaded region) indicate that the neuron shows no net bias to either of the stimuli in the pair. B: LMAN neurons in sequentially tutored adult birds exhibited heightened selectivity for the BOSAdult vs. other stimuli. The mean d' values for all responsive neurons (n = 84) were similar for the BOSAdult vs. TS1, BOSAdult, and CON, but significantly less for BOSAdult–TS2 (asterisk indicates P < 0.05, ANOVA test).

View larger version (16K): [in this window] [in a new window]   FIG. 12. Developmental comparison of LMAN auditory responses in juvenile and adult sequentially tutored birds reveals loss of responsiveness for early generated songs and a loss of responsiveness to the tutor model on which these juvenile songs are based. The mean RS (spikes/s) to a given song at PHD 60 before exposure to the second tutor (open circles) and in adults after exposure to the second tutor (filled squares). Significant decreases in responsiveness to the BOSPHD60 and TS1 occurred over development. Also note the equivalent RS values for the BOS and the most recent tutor song at the 2 developmental time points. "REV" refers to reverse playback of the relevant (i.e., age-matched) BOS. P values obtained using a Welch's t-test (BOSPHD60) and Student's t-test (TS1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

Behavior of sequentially tutored birds

The sequential tutoring approach provided juveniles with 2 tutors, affording them an opportunity to copy 2 song models. Birds tutored in this manner could copy from each tutor, overwriting the copy of the first model as the bird mimicked the second tutor. Just before the introduction of the second tutor, the BOS exhibited features indicative of partial copying from TS1. This early copying was most apparent in juvenile zebra finches first exposed to a Bengalese finch tutor, but could also be seen in some juvenile birds raised initially by another conspecific. In contrast, sequentially tutored adults produced songs with copied syllables derived largely from the second and not the first tutor, although more subtle effects of TS1 may have persisted in the adult song of some birds, as indicated by the elevation in Sound Analysis similarity scores >50%. Certain songbird species, including swamp and song sparrows, copy from several different tutors, then undergo selective attrition, retaining only a subset of these vocalizations to form a more limited suite of adult song types (Marler and Peters 1981, 1982a; Nordby et al. 2000). In their capacity to mimic multiple vocal models and retain only a subset of these juvenile imitations in the adult vocal repertoire, sequentially tutored zebra finches roughly resemble these other songbirds. One issue emerging from our results is whether the serial "bilingual" birds we studied retain permanent auditory or motor memories of the songs they transiently learn. In swamp sparrows, song material deleted in the first year can be reexpressed during periods of plastic song production in subsequent springs, implying retention of auditory or motor memories of the "lost" songs (Marler and Peters 1982b). If such auditory memories persist, then one may wonder whether early experience, especially of other species' songs, extends perceptual performance, as for humans with early experience of a language distinct from their native tongue (Tees and Werker 1984).

Sequential tutoring may more closely approximate some aspects of zebra finch song learning in the wild than protocols using only a single tutor. Juvenile zebra finches become independent about PHD 35 (i.e., slightly after exposure to TS1 in these experiments), potentially exposing them to the songs of other conspecifics in addition to their father's song (Immelmann 1969; ^</