| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Department of Biology and 2Neuroscience Graduate Group, University of Pennsylvania, Philadelphia, Pennsylvania
Submitted 26 September 2005; accepted in final form 26 July 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
). Although most studies have used species that sing a single adult song, study of other species possessing multiple songs remains essential for full evaluation of models of vocal learning, perception, and production (Del Negro et al. 2005; Nealen and Schmidt 2002). Much of the theoretical work in songbird vocal learning makes use of a single "bird's own song" (BOS) as an immutable template or target, but to learn and use multiple vocalizations may be the more common strategy (e.g., Kroodsma and Baylis 1982).
One site of considerable interest within the avian song neural system, nucleus HVC (used here as the proper name), lies at the confluence of auditory and motor streams (McCasland and Konishi 1981; Nottebohm et al. 1976; Fig. 1). HVC neurons exhibit complex auditory properties (Lewicki and Arthur 1996; Theunissen and Doupe 1998) and may function in the auditory-motor transformation necessary for song learning and maintenance (Troyer and Doupe 2000). Under anesthesia, HVC auditory responses in several species are consistently selective for BOS over conspecific (CON) song (Cardin and Schmidt 2003; Janata and Margoliash 1999; Margoliash 1986; Mooney et al. 2001). Mooney et al. (2001) provided the first description of song repertoire representation within nucleus HVC of the anesthetized swamp sparrow Melospiza georgiana, demonstrating that HVC neuronal subtypes have distinct auditory properties and may exhibit single-syllable selectivity. Given that swamp sparrow songs consist of repetitions of a single syllable, it is difficult to predict how HVC neurons would respond to multiple songs consisting of variable sequences of different syllables. It is also unclear why the robust and selective auditory responses observed under anesthesia in the zebra finch and other species have not been observed in awake animals (Cardin and Schmidt 2003, 2004; George et al. 2005).
|
; Fig. 2B). Neighboring males may song type-match, singing nearly identical copies of subsets of their song repertoires (Beecher et al. 2000). In contrast to the colonially breeding zebra finch, song sparrows are strongly territorial, are highly responsive to conspecific song, and can discriminate among individuals and among syllable types on the basis of song (Nelson 1987; Stoddard et al. 1991).
|
). Using measures of the spatial reliability of auditory responses across multiple HVC recording sites, we show that auditory responses are more spatially consistent in awake than in anesthetized animals, a finding that suggests that the neural basis for auditory selectivity within nucleus HVC is not entirely conserved between the awake and anesthetized conditions. |
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Adult song sparrows were located on breeding territories in natural habitats in southeastern Pennsylvania. Spontaneous song was recorded from territorial males by portable analog recording equipment (Sony PBR-330 parabolic reflector, Optimus 33–3014 microphone, Marantz PMD-430 cassette recorder). Song was recorded from both focal and neighboring males (Fig. 2A). Although it was not possible to exhaustively sample the entire song repertoire of individuals (bird's own repertoire [BOR]), song recording for several hours was sufficient to identify a frequently used subset of an individual's song repertoire, consisting of multiple, different BOS (bird's own song) song types. When possible, song was recorded from birds that had been color banded for individual identification; song was collected from unbanded males only when individuals could reliably be followed on their territories. Focal males were then captured by mist nets and brought to the laboratory. Captured birds were held in individual cages in a common aviary. State and federal permits for the capture and use of wild birds were in hand. All methods were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania.
Laboratory sound recording and sound stimulus preparation
Before neural recordings, individual male song sparrows were housed for several days in sound-attenuation chambers (Acoustic Systems, Austin, TX) for computer-based recording of spontaneous songs. A minimum of four "bird's own song" exemplars representing unique song types (BOS1–BOS4; Fig. 2B) were chosen from the field and laboratory song recordings available for each individual for use as auditory stimuli during neural experiments.
Song stimuli used for playback were digitized at 40 kHz and high-pass filtered at 300 Hz. Four acoustic variants of individual BOS sound stimuli were constructed for use as test stimuli, including BOS (in its original, forward orientation) and REV (whole-song reversed) sound stimuli. Other acoustic variants included song stimuli whose syllables were presented in the original sequence but in individually reverse orientation (R-SYL) and stimuli consisting of the original song syllables played in correct orientation but in reverse-order sequence (R-ORD). These latter classes of stimuli were used to assess the effects of syllable sequencing and syllable context in eliciting auditory responses. If, for example, the acoustic features of individual syllables contribute most to HVC auditory responsiveness, R-ORD sound stimuli should elicit robust auditory responses. If, however, the manner in which acoustic spectra evolve over the time course of a sound stimulus is critical to auditory responsiveness, R-SYL sound stimuli would be expected to provide auditory drive similar to that of natural BOS stimuli. Short segments of a white-noise (WN) stimulus and a heterospecific (HET; zebra finch Taeniopygia guttatta) song stimulus were used during some experiments to measure the degree of nonspecific auditory responsiveness. In addition to the spontaneous neural data collected before each auditory presentation, during some experiments neural data were collected during experimental trials containing no auditory stimulus (BLANK) that were interleaved among auditory test stimuli. Auditory trials involving BLANK, WN, conspecific (CON), and HET stimuli were excluded from present analyses.
Experimental procedure
ACUTE EXPERIMENTS. Animals were removed from food/water for 1–2 h before administration of either diazepam (10 mg/kg) or a combination of ketamine/xylazine (35/7 mg/kg), injected intramuscularly. After drug onset, subjects were placed into a protective surgical jacket heated to 35°C. Scalp feathers were plucked; topical antiseptic (Betadine, Purdue Frederick, Stamford, CT) and anesthetic (2% lidocaine, Copley Pharmaceutical, Canton, MA) were administered to the scalp. Subjects were then placed into a stereotaxic apparatus. The scalp was split and retracted, and small ink marks were made on the skull surface above nucleus HVC, as located by stereotaxic coordinates. A small post was cemented to the anterior skull with dental acrylic (Denstsply, York, PA) and then used to support the animal's head, leaving the auditory canals unobstructed.
Recordings were made unilaterally in nucleus HVC (Fig. 1A); both left and right hemispheres were sampled. A craniotomy (150–300 µm) was made over the desired recording location and the dura was cut with a sharpened tungsten wire. Extracellular recordings were made using commercial tungsten electrodes (2–5 M
, A-M Systems, Carlsborg, WA) or glass micropipettes (A-M Systems) of tip diameter 8–15 µm filled with 4 M K-acetate. In several subjects, recordings were performed by cell-attached recording, providing high-quality single-neuron isolation. Neural signals were amplified 500x and filtered between 300 and 10,000 Hz (HS4 headstage and DB4 Bio-amp controller, Tucker-Davis Technologies, Alachua, FL; Brownlee 410 amplifier, Brownlee Precision, San Jose, CA). Neural signals were monitored on-line by oscilloscope (Hameg Instruments, Frankfurt, Germany) and sound monitor (Grass-Telefactor, West Warwick, RI).
Correct neural recording location was achieved by stereotaxic electrode placement, confirmed by diagnostic spontaneous neural activity. Recording locations in initial study animals were confirmed by histological examination; all were confined to within the boundaries of nucleus HVC, as defined by cresyl violet stain for Nissl substance. Once neural recording was achieved, sound stimuli were presented by passive playback in an interleaved fashion, with variable intertrial intervals (5–15 s). Sound stimuli were normalized to achieve maximum amplitude of 70–75 dB at the animal. Auditory presentation and concomitant neural data were digitized at 20 kHz and recorded by custom software (A. Leonardo, Cal Tech) for off-line analysis.
At each recording location, a small number (1–18, mean: 3.92) of the available sound stimuli for each animal was presented, with each stimulus presented between 5 and 100 (mean: 31) times. After a sound presentation block, the recording electrode was moved a minimum of 50 µm to a new recording location before further presentation of sound stimuli.
After 2–3 h of auditory stimulation, the recording electrode was withdrawn. The skull craniotomy was filled with GelFoam (Pharmacia/Upjohn, Kalamazoo, MI) and the scalp closed with sutures. The animal was removed from the surgical apparatus and allowed to recover from anesthesia in an undisturbed, heated cage. Recovery from anesthesia was rapid; animals typically resumed perching and feeding within 1–2 h. Subjects were allowed to recover a minimum of 1 wk before further neural recording.
CHRONIC EXPERIMENTS.
Subjects were prepared for neural recording as described above. Several animals received bilateral, fixed electrodes in the following manner. After stereotaxic location of left and right HVC, a single blunt-end recording electrode (0.5–1.5 M) made from Formvar-insulated nichrome wire (A-M Systems) was placed into each left and right HVC, as described previously (Schmidt 2003). Electrode leads were connected to a microconnector (Omnetics, Minneapolis, MN) cemented to the animal's skull. Other animals received unilateral tungsten electrodes (from one to three in number), movable by custom-built manual or motorized microdrives (Fee and Leonardo 2001).
After a 24-h recovery period, animals were connected by a wire tether to a custom recording chamber, which allowed full range of motion and continuous neural recording. Subjects exhibited all normal behaviors while in the recording chamber, including perching, feeding, calling, and singing. The recording chamber itself was housed within a sound-attenuation chamber containing a speaker over which sound stimuli were presented (as above) for assessment of neural responses during auditory presentation in awake animals. A camera and microphone were used to remotely monitor the animals during these experiments; at no time were animals observed to sleep during sound presentation.
Data processing and analysis: extraction of spike events and general methods
Neural recordings were processed by custom software in the MatLab environment (The MathWorks, Natick, MA). All statistical analyses were performed using SAS ver. 8.00 (SAS Institute, Cary, NC) and Microsoft Excel (version 2002). Data values are presented as means ± SE. Distribution central tendencies were compared using ANOVA (F-statistic reported) or by t-test (t-statistic reported). Comparisons of t-tests were paired by recording location where reported and are unpaired otherwise. Before t-test computation, distributions were evaluated for equivalence of variance. Contingencies were evaluated using sign tests (M value reported).
Neural traces dominated by the activity of one or a few individual neurons were subject to template-based spike detection and sorting (Spike2 software, version 5, CED, Cambridge, UK). Other, multiunit neural traces were subjected to thresholding spike detection. Single- and multiunit recordings are combined in analyses except where indicated. All spike event times were binned at 10-ms intervals for analyses. For all repetitions of a given sound stimulus at one recording location, neural activity event times were summarized across trials into a peristimulus time histogram (PSTH; see Fig. 3). PSTH bin values were normalized to the number of auditory trials in each experimental block and thus ranged from 0 (no neural activity over a given 10-ms interval during any of the auditory presentations of a sound stimulus) to 1 (neural activity in a given 10-ms interval during every presentation of a sound stimulus). The vast majority of PSTH bin values reflect values between these two extremes. For each of the three phases of an experiment (before [duration 1.0–1.5 s], during [duration 2–4 s], and after [duration 2–4 s] auditory stimulation), the minimum, maximum, mean, and variance were calculated from PSTH bin values. Minimum, maximum, and mean of PSTH bin values reflect the overall levels of neural activity, whereas variance in PSTH bin values is a measure of the modulation of excitation during auditory presentation; high variance suggests sensitivity to temporal structure in acoustic stimuli.
|
 |
)
 |
2 refer to the strength and variance, respectively, of auditory responses to song stimuli A and B.
No selectivity (e.g., equivalent responses to two stimuli) would result in a selectivity index of 0 (Green and Swets 1966; Janata and Margoliash 1999). Given that this metric is symmetric about 0 (Janata and Margoliash 1999), the degree to which observed indices reflect consistent selectivity was tested by comparing the distributions of observed selectivity indices to 0.
Stimulus ensemble reliability measurements
Multiple BOS stimuli from any one bird's repertoire were used as test stimuli at multiple recording locations within HVC in that individual. As described above, the size of the auditory response evoked by any one test stimulus was measured as the elevation in neural activity over prestimulus levels (RS value). To test the degree to which auditory representations in nucleus HVC are global in nature (Sutter and Margoliash 1994), the magnitude of response evoked by a given stimulus was compared across the recording locations at which a stimulus was tested.
To assess the uniformity of auditory responses evoked by test stimuli, the set of auditory responses evoked by presentation of an ensemble of BOS songs was compared across recording locations within individuals. At any one recording location (j), the BOS stimuli tested (BOS1–i) were ranked (rj) according to the magnitude of the auditory response (RS value) evoked by each stimulus. The order of this ranking was then compared across each location at which the stimulus ensemble was tested. If recording locations across HVC represent BOS auditory stimuli in the same way, the relative ranking of responses evoked by a set of test stimuli would repeat across recording sites. For example, consider an ensemble of three test stimuli (i: BOS1–3) tested at each of four recording locations (j1–4). At recording location j1, the auditory responses (measured as RS) evoked by these stimuli could be ranked (r1) as follows
 |
|
|
 |
 |
 |
|
|
 |
 |
 |
 |
|
The lower reliability scores obtained in this example reflect the greater variability in relative stimulus ranking across recording locations. For both awake and anesthetized data, reliability scores were computed for any ensemble of test stimuli used at multiple locations within an individual HVC. To assess the utility of reliability measures, the set of observed reliability scores was compared with those obtained after a uniform shuffling of stimulus ranks and with those obtained from a model data set of equal size that exhibited perfect repeatability of stimulus ranks. To assess the potential for state differences in auditory processing, test stimuli reliability scores were compared between recordings collected in the awake and anesthetized states.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
After 1–2 days of acclimation, male song sparrows originally observed and recorded on territories in their natural habitats sang spontaneously in laboratory acoustic-attenuation chambers. Song production varied broadly, ranging from several hundred to several thousand songs sampled per individual over multiple days of recording. All males recorded in the laboratory sang most often in the few hours after the onset of the daylight portion of the light cycle, suggesting that study individuals retained their normal diurnal singing pattern.
HVC auditory responses are elicited in both awake and anesthetized song sparrows
Auditory responses in nucleus HVC were apparent as an elevation in single- or multiunit spike rate during auditory stimulation. In both awake (Fig. 3A) and anesthetized (Fig. 3B) animals, auditory responsiveness was broad, with most song stimuli eliciting an elevation in HVC activity throughout the duration of auditory playback. RS magnitude to forward-orientation self-song (BOS) playback did not differ between the two sedative mixtures used [ketamine/xylazine vs. diazepam; F(1,61) = 1.66, NS]; data representing use of these drugs were therefore pooled in subsequent analyses.
Multiple BOS song types from a bird's own repertoire (BOR) were used as auditory playback stimuli at each recording location. In awake animals, presentation of BOS song stimuli caused, on average, a 44% elevation (e.g., mean RS = 1.44) in neural activity over prestimulus levels (n = 32 recording sites in five birds; Fig. 3C). After the end of auditory stimulation, HVC activity returned to prestimulation levels (average of 1% greater than prestimulus levels, NS). In anesthetized animals, presentation of BOS stimuli evoked, on average, a 103% increase in HVC activity over prestimulus levels (n = 64 recording sites in 11 birds; mean RS = 2.03; Fig. 3C). In contrast to awake animals, this excitatory response during auditory stimulation was typically followed by a period of activity suppression, to a level significantly less than that before auditory stimulation (sign test, M = 13.0, P < 0.005). This activity suppression, reported previously in both intracellular (Lewicki 1996) and extracellular recordings (Volman 1996), lasted for several seconds and was of small but consistent magnitude (mean: –6.2% after forward-orientation BOS stimuli). Overall, the magnitude of elevation of HVC activity during auditory playback was greater in anesthetized than in awake animals (t-test, t1,94 = 2.19, P < 0.05; Fig. 3C).
Increases in HVC activity during auditory stimulation were accompanied by increased variability of activity (measured as the variance over 10-ms intervals across PSTH; Cardin and Schmidt 2003). In awake animals, the 44% increase in overall activity from prestimulus to during-stimulus periods was associated with a 354% increase in the temporal variance of activity, whereas the 103% elevation of overall activity during stimulation under anesthesia was accompanied by a similar 353% increase in the variability of neural activity during auditory stimulation. The fact that presentation of sound causes the mean and variance of HVC neural activity to increase by different amounts suggests that, in both animal states, auditorially evoked excitation in this brain region is not achieved by a uniform "gain control" over spontaneous neural activity levels.
HVC auditory responses were selective for BOS over REV sound stimuli
Presentation of BOS auditory stimuli in normal, forward orientation reliably evoked robust auditory responses in nucleus HVC of both awake (Fig. 4A) and anesthetized animals. To explore the acoustic basis for the generation of these responses, we compared the responses elicited by artificial temporal and spectral variants of normal BOS stimuli, including whole-song reversed (REV) sound stimuli. Both BOS and REV stimuli reliably evoked HVC auditory responses, of broadly variable magnitude. Across all animals and recording sites, BOS auditory responses were greater in overall magnitude (whole-song RS value) than REV responses, a difference that was significant in awake [mean ± SE: 1.47 ± 0.10 vs. 1.16 ± 0.04; F(1,50) = 5.42, P < 0.03] but not anesthetized recordings [2.07 ± 0.25 vs. 1.67 ± 0.18; F(1,78) = 0.67, NS; Fig. 4B]. Nonetheless, assessments of auditory selectivity using the d' measure, which make use of both the magnitude and trial-to-trial variability of responses in paired comparisons, revealed consistent selectivity of BOS over REV responses in both awake (mean d' ± SE: 0.34 ± 0.08) and anesthetized (mean d' ± SE: 0.41 ± 0.12) birds (Fig. 4C).
|
In both awake and anesthetized conditions, BOS song stimuli evoked auditory responses of greater magnitude (as measured by RS) than did use of acoustic variants (R-ORD, R-SYLL, REV stimuli; see METHODS) of BOS stimuli. In both animal states, R-ORD stimuli evoked auditory responses of greater RS magnitude than did REV or R-SYLL stimuli (Fig. 5). BOS and R-ORD stimuli share normal orientation of individual syllable acoustic structure (while differing in syllable sequence order), whereas the stimulus categories that evoked the smallest auditory responses (REV or R-SYL) present individual syllables in reverse temporal form. These findings suggest that the acoustic features of individual syllables contribute more strongly to HVC auditory responsiveness than does temporal order of song-level spectral content.
|
The song types that make up an individual song sparrow's BOR vary broadly in spectral content (Fig. 2B). To assess the stimulus-specific contribution to the variability seen in BOS auditory responses, RS values from passive presentation of multiple BOS stimuli were compared within recording locations. Multiple BOS stimuli were presented while recording at a total of (nonsimultaneous) 19 unique nucleus HVC recording locations in awake animals and at a total of 59 recording locations in the HVC of anesthetized animals. At any given recording location, the magnitude of evoked auditory response varied among the BOS stimuli tested, in both awake and anesthetized animals (Fig. 6). Figure 6A depicts a single example of this type of variation. Although each of the four BOS stimuli tested provided reliable auditory drive at this recording location (all RS >1), the degree of excitation provided varied among the different BOS stimuli presented (compare RS for BOS1 vs. BOS3), and for only one stimulus (BOS1) was this the recording location at which the sound stimulus evoked its greatest auditory response.
|
As used here, RS values represent an overall, "whole-song" level of auditory responsiveness. Because many auditory responses included phasic elements (e.g., Fig. 4), it was necessary to verify that this whole-song measure accurately reflected more transient elements of neural auditory responsiveness. To do so, auditory responses were scored to identify the smallest (RSmin) and largest (RSmax) auditory responses for each recording location. Then, responses gauged to be the smallest and largest at each recording location on the basis of whole-song RS values were compared with respect to measures of auditory excitation calculated from 10-ms PSTH bin values. By multiple measures of comparison (mean, maximum, and variance of neural excitation during auditory presentation, as measured over 10-ms PSTH bin values), auditory responses assessed by whole-song RS as being "largest" or "smallest" within recording locations were statistically separable in both awake (paired t-test, 19 recording locations: t1,18 = 3.13–4.46, all P < 0.01) and anesthetized recordings (59 recording locations: t1,58 = 6.12–6.62, all P < 0.0001). These results suggest that whole-song RS rank values accurately represent the relative excitation evoked by BOS song stimuli, despite the existence of rapid modulations to the neural excitation provided by individual song stimuli.
Measures of RSmin and RSmax were then combined to provide an estimate of the neural dynamic range (RSrange = RSmax/RSmin) for BOS auditory responses within individual HVC recording locations. Awake and anesthetized recording locations differed broadly in overall auditory responsiveness; recording location RSmin, RSmax, and RSrange were greater in anesthetized than in awake animals (RSmin: t1,76 = 2.97, P < 0.005; RSmax: t1,76 = 3.11, P < 0.005; RSrange: t1,76 = 1.66, P = 0.05). In awake animals, HVC within-location RSrange averaged 1.3-fold (average RSmin: 1.19; average RSmax: 1.55; Fig. 6B). Under anesthesia, HVC within-location BOS auditory dynamic range averaged 1.4-fold (average RSmin: 1.53; average RSmax: 2.19). This within-location variability in auditory responsiveness to different songs of the BOR suggests that individual neurons and/or recording sites are specifically tuned to acoustic features present in a subset of songs in the BOR.
Auditory tuning to BOS: variation among HVC recording locations
HVC auditory representation of BOS in the single-songed zebra finch is known to be broadly similar across the entire extent of the nucleus under anesthesia (Sutter and Margoliash 1994). This finding suggests a uniformity of auditory tuning across HVC to match the spectral characteristics of an individual's BOS. However, for each song sparrow individual studied in these experiments, the multiple BOS songs used as test stimuli represent only a subset of the song types within each individual's BOR. BOS song types vary widely in spectral characteristics (Fig. 2B), yet all BOS sound stimuli reliably evoked auditory responses in nucleus HVC in both anesthetized and awake song sparrows. This suggests that individual song sparrow BOS stimuli may drive some HVC recording locations to a greater degree than others.
To test whether HVC auditory responsiveness varies spatially, we compared the RS values obtained from passive presentation of multiple songs from an individual bird's repertoire at multiple locations within that individual's nucleus HVC. The ability of a given sound stimulus to elicit auditory responses in nucleus HVC varied broadly with location in the nucleus. As an example, the RS values generated by presentation of a single BOS stimulus at four different recording locations within nucleus HVC are presented in Fig. 7A. This stimulus generated the maximum observed RS (top tick mark in each bar) at only one of the four sites tested (location #4), and was the stimulus providing the least auditory drive (bottom tick mark in each bar) at none of the four recording locations.
|
), does not obtain in the song sparrow. Estimates for stimulus RSmax did not increase with the number of recording locations sampled in either awake [F(1,11) = 0.95, NS] or anesthetized [F(1,35) = 0.03, NS] states, suggesting an adequate sampling of among-recording location variation in RSBOS. Auditory responses reflect a combination of stimulus properties and local tuning
Both sound stimulus acoustic properties and local auditory tuning within nucleus HVC contribute to auditory responsiveness (Fig. 8). The combination of location-specific and stimulus-specific variability slightly exceeds that observed in both awake (RSBOS observed: 1.8x; combined estimate: 1.5x x 1.3x = 1.95x) and anesthetized (RSBOS observed: 3.4x; combined estimate: 2.5x x 1.4x = 3.50x) recordings, suggesting a slight amount of covariance between stimulus and location-specific measurements. In both the awake and anesthetized conditions, the relative contribution of sound stimulus characteristics to the variability of auditory response magnitudes is less than that contributed by the auditory tuning of the given recording location (awake: RSrange of location = 1.5-fold, RSrange of stimulus = 1.3-fold; anesthetized: RSrange of location = 2.5-fold, RSrange of stimulus = 1.4-fold; Fig. 8). Thus song repertoire auditory representation within HVC is neither completely spatially distributed nor globally uniform. Given that song types may share common syllables (Podos et al. 1992), local HVC tuning to individual syllable types, as described by Mooney et al. (2001), may be sufficient to generate the representation we describe here.
|
The complexity and variability of song sparrow song types, in conjunction with the spatial variability of HVC auditory tuning, suggest a more complex auditory representation of self-song than has been described for the zebra finch. To assess the spatial organization of auditory representation within this brain region, we measured the repeatability with which stimulus ensembles (groups of two to four unique BOS song types) were represented across recording sites. For those recording sites at which two or more BOS forward-orientation songs were used as auditory playback stimuli, the test stimuli were ranked according to the magnitude of evoked RS. The RS ranks of each member of the stimulus ensemble were then compared across each recording site at which the ensemble was tested. These ranks were then used to produce reliability scores for each sound stimulus that represent the consistency of auditory responses across HVC recording sites (see METHODS).
To demonstrate the utility of ranking the members of stimulus ensembles in this way, stimulus ranks achieved from presentation of two equally sized stimulus sets (one each from awake and anesthetized HVC recordings, from different birds) are depicted in Fig. 9A. Stimulus ensembles used in these examples consisted of four BOS song stimuli, tested at each of five recording locations within nucleus HVC. At each recording location, the BOS stimuli were ranked with respect to the relative magnitude of the evoked auditory response (e.g., ranked within recording location by RS value). As depicted in Fig. 9A, variation across recording locations in the relative size of the evoked auditory response to a given stimulus is substantial in both awake and anesthetized recordings. However, this variation is not structured equivalently under awake and anesthetized conditions. Figure 9B depicts the stimulus ranks and reliability scores achieved by the stimuli represented in Fig. 9A. The four sound stimuli tested in the awake animal achieved reliability scores from 0.36 to 1.00, and no single sound stimulus achieved all possible rank positions within the ensemble. Reliability scores for the four sound stimuli tested under anesthesia ranged from 0.28 to 0.52, and two of the four sound stimuli achieved both the lowest and highest possible rank positions within the ensemble. Ensemble mean reliability is lower in the anesthetized (0.38) than in the awake recording (0.60), suggesting that stimulus rank repeatability varies with animal state.
|
To evaluate the likelihood of observed reliability scores, scores obtained during recordings in both awake and anesthetized conditions were compared with two permutations of the observed data. First, observed scores were compared with data sets of matching size that represented perfect stimulus reliability (e.g., all reliability scores = 1). Second, observed scores were compared with those obtained from a shuffled version of the observed data (reliability scores computed after random ordering of stimulus rankings within each recording location). Stimulus ensemble reliability scores differed considerably across treatments [F(5,48) = 21.45, P < 0.001]. The reliability scores obtained in both awake and anesthetized conditions were significantly lower than 1 (="perfect" reliability; statistics not shown). Reliability scores from nucleus HVC recordings in awake animals were greater than those obtained from a random shuffling of the observed data (Fig. 9C; P < 0.05), whereas scores obtained under anesthesia did not differ from a shuffled version of scores obtained during anesthetized recordings (Fig. 9C; NS).
The example described above and depicted in Fig. 9, A and B is representative of the overall difference in stimulus reliability seen in awake and anesthetized animals. Reliability scores summarized across all members of sound stimulus ensembles reveal than stimulus reliability is significantly greater in awake nucleus HVC than under anesthesia (Fig. 9C; t1,53 = 1.92, P < 0.05). High- and low-ranking stimuli did not differ in overall power spectra (not shown), which suggests that response magnitudes reflect a matching of stimulus and local neuronal tuning characteristics, rather than an overall bias for specific spectral features. These comparisons demonstrate that the spatial organization of stimulus auditory representation within HVC differs between animal states, a finding that suggests caution in the interpretation of awake animal neural phenomena on the basis of data collected under anesthesia. These results also suggest that individual song-level selectivity measures alone may be inadequate to fully describe the representation of song repertoires drawn from a large, shared pool of song syllables.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Auditory excitation within nucleus HVC
Presentation of BOS song stimuli reliably evoked excitation within nucleus HVC in both awake and anesthetized animals. As reported above (see RESULTS), the magnitude of BOS-evoked excitation (expressed as percentage increase over spontaneous activity) was greater in anesthetized than in awake animals, and yet the variance in auditory neural activity increased by an equivalent amount in both animal states. That the temporal variability of neural activity during auditory stimulation changes equivalently between states may reflect the basic temporal structure of song sparrow song stimuli, because all stimuli consisted of distinct syllables separated by silent intervals 
0.3 s in duration (Fig. 2B). Alternatively, intrinsic dynamics of the HVC excitatory response may act in common across both animal states to provide this common structure to the increase in network excitation; direct tests of either possibility have not been performed.
Species differences in HVC auditory responsiveness
HVC auditory processing in song sparrows is notably distinct from that of the zebra finch in the relatively modest differences in auditory response magnitudes between awake and anesthetized conditions and the presence of robust and stable auditory responsiveness in the awake animal. Overall, the fact that robust auditory selectivity in HVC is apparent in awake song sparrow, but not zebra finch, suggests a fundamentally different utility to the audition of song in these species. These species differ strongly in both their territorial use of song and the level of song sharing among "neighbors" (Arcese et al. 2003; Zann 1996); whether the neural auditory differences reported here relate to either or both of these behavioral aspects of song use is yet to be determined. Behavioral and/or ecological differences among species may explain some of the differences because passive audition of a BOS song is an unnatural event for a zebra finch, but represents a potentially relevant signal from a song type-matching song sparrow conspecific. Passive presentation of song stimuli is insufficient; however, to fully study interindividual recognition and signaling in this (and other) species, as visual signals, physical interaction, and nonsong vocalizations also facilitate intraspecific communication. Further tests, incorporating behavioral interactions in ethologically relevant settings, will be necessary to determine whether HVC auditory responses are associated with the robust behavioral response to conspecific song in this species.
Song sparrow HVC auditory responses are consistently selective for BOS song
Unlike the zebra finch (Cardin and Schmidt 2003), the degree of song sparrow HVC BOS-REV selectivity did not differ between states (t1,35 = 0.47, NS). In the zebra finch, d' auditory selectivity values for awake zebra finch HVC averaged 0.05 ± 3.5 (Cardin and Schmidt 2003) and were broadly distributed around 0, demonstrating an overall lack of auditory selectivity in HVC of awake birds. Although the overall magnitude of song sparrow HVC BOS-REV selectivity is less than that reported from studies in anesthetized zebra finch nucleus HVC (Cardin and Schmidt 2003; Lewicki and Arthur 1996; Theunissen and Doupe 1998; Theunissen et al. 2004), neural population-level BOS-REV selectivity in HVC is apparent in both awake and anesthetized song sparrows.
Overall, song sparrow HVC selectivity indices for BOS versus REV (Fig. 4) are modest in comparison to anesthetized zebra finch HVC and greater than those reported for awake zebra finch HVC. Relatively few song sparrow HVC recording locations meet the threshold of d' values of >0.5 or 1.0, used to categorize anesthetized zebra finch recording locations as "selective" or "nonselective" (Theunissen and Doupe 1998). However, the use of a single, positive threshold value in this way is arbitrary; no use of such a threshold is indicated in the original derivation of this metric (Green and Swets 1966). As described by Janata and Margoliash (1999), the d' metric is symmetric about 0. Selectivity indices representing single recording locations are free to vary in either the positive or negative direction, which suggests that a population-level description of auditory selectivity is best revealed by comparing observed indices to 0. Although the magnitudes of selectivity indices observed here are smaller than those commonly reported for the zebra finch, the majority of indices are >0 (Fig. 4). Thus auditory selectivity may be less extreme than reported for the zebra finch, but is apparent nonetheless. Interpretation of auditory selectivity remains difficult because we cannot assess how the neural utility of an HVC BOS auditory response differs from that of a different song type, or of conspecific song. Similarly, we cannot, at present, relate HVC neural responses to associated behavioral responses; it is known from other systems that very modest neural sensory response differences may translate nonlinearly into sharply demarcated behavioral decisions (e.g., Asaad et al. 2000; Kim and Shadlen 1999).
Organization of BOS auditory representation within HVC
HVC auditory responses to the single song of the zebra finch are relatively consistent across multiple recording locations within anesthetized nucleus HVC (Sutter and Margoliash 1994). However, the auditory representation of song repertoires may require something other than a global strategy of single-BOS auditory representation because song sparrow HVC auditory responsiveness differs from that of the zebra finch in several important ways. First, auditory responses to BOR songs in song sparrows are not individually or collectively uniform across recording sites under anesthesia (Fig. 7). In anesthetized song sparrow HVC, the magnitude of auditory responses to BOR stimuli vary more than threefold among recording locations and stimulus types (Fig. 8), with both sound stimulus and local tuning contributing substantially to auditory response variation. That individual recording sites vary in the degree to which they can be driven by a given sound stimulus suggests that the neural valence of a particular stimulus is dependent on recording location. Thus song repertoire auditory representation in this species is global, or nonspecific, only in the sense of broad auditory responsiveness across nucleus HVC. Each of an individual's BOR song types reliably evokes HVC auditory responses, although the stimulus that does so best varies by recording location. No single BOR song type, or "super-stimulus," could be identified in each animal as consistently providing the greatest degree of auditory drive.
Second, auditory response reliability indices (Fig. 9) suggest an organized spatial structure to awake, but not anesthetized, HVC auditory responsiveness in this species, a finding that suggests that the network dynamics that generate auditorially evoked excitation within HVC are not consistent across animal states. How these species differences relate to the shared-syllable organization of the multiple songs of the song sparrow repertoire remains to be determined. It is possible that syllables, rather than song types, are discretely represented within nucleus HVC (Mooney et al. 2001), and syllable sharing among song sparrow song types itself lends consistency to auditory representation across this brain region. The mechanism by which this relationship would change between the awake and anesthetized states remains elusive, although the reduced repeatability of stimulus ensembles under anesthesia is consistent with the relatively reduced selectivity and enhanced auditory responsiveness that characterize HVC neural recordings in this state.
In conclusion, selective and acoustic-feature–specific auditory responses are a hallmark of the avian song system, yet their existence remains enigmatic because they have rarely been described in awake, behaving animals. We demonstrate here that robust and selective auditory responses exist in the avian song system nucleus HVC in awake animals, in a species with known behavioral use of song in intraspecific communication. We further demonstrate that auditory responsiveness results from a combination of local neuronal tuning with sound stimulus acoustic features and provide evidence for a state-dependent, spatial organization to the auditory representation of song repertoires within this brain region. Current models of song learning and maintenance posit a role for the passage of auditory information from HVC into the basal ganglia-like anterior forebrain and the present study provides strong evidence for the availability of auditory information in the nucleus HVC of awake animals. Although HVC auditory responsiveness during passive song presentation may differ functionally from the real-time auditory feedback potentially available during singing, the BOS-specific auditory responses described here potentially reflect the importance of such feedback for song learning and maintenance. These studies provide a gateway for further investigation of song perception in awake, behaving animals and point to the need for careful consideration of the behavioral use of song in the design of experiments on the neural basis of song processing.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Present address of P. M. Nealen: Department of Biology, Indiana University of Pennsylvania, 114 Weyandt Hall, 975 Oakland Avenue, Indiana, PA 15705–1001.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: P. M. Nealen, Department of Biology, Indiana University of Pennsylvania, 114 Weyandt Hall, 975 Oakland Avenue, Indiana, PA 15705-1001 (E-mail: pnealen{at}iup.edu)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Asaad WF, Rainer G, and Miller EK. Task-specific neural activity in the primate prefrontal cortex. J Neurophysiol 84: 451–459, 2000.
Beecher MD, Campbell SE, Burt JM, Hill CE, and Nordby JC. Song-type matching between neighbouring song sparrows. Anim Behav 59: 21–27, 2000.[CrossRef][ISI][Medline]
Brenowitz EA and Beecher MD. Song learning in birds: diversity and plasticity, opportunities and challenges. Trends Neurosci 28: 127–132, 2005.[CrossRef][ISI][Medline]
Cardin JA and Schmidt MF. Song system auditory responses are stable and highly tuned during sedation, rapidly modulated and unselective during wakefulness, and suppressed by arousal. J Neurophysiol 90: 2884–2899, 2003.
Cardin JA and Schmidt MF. Auditory responses in multiple sensorimotor song system nuclei are co-modulated by behavioral state. J Neurophysiol 91: 2148–2163, 2004.
) and anesthetized animals (
). Distribution means are indicated by a horizontal bar in each distribution. Inset numbers: number of recording locations at which each category of stimulus was tested.
2 different BOS song stimuli were tested, the RSmin and RSmax were scored for that location; population average RSmin and RSmax are plotted as horizontal lines. Data collected from each individual bird are grouped together here for presentation; animal numbers and inset lines at bottoms of panels span the recording locations sampled from each individual bird.
