A highly developed auditory network supports auditory-vocal behavior in songbirds. The core of the auditory processing system consists of anatomical areas named Field L, NCM (caudomedial nidopallium), and CM (caudomedial mesopalium) (Vates et al., 1996) (Figure 1c). These areas and other associated auditory areas are directly and indirectly connected with the song motor pathway (Vates et al., 1996; Mandelblat-Cerf et al., 2013). Field L, the primary thalamorecipient area, is composed of four different sub-regions (L2a, L2b, L1, and L3) that are reciprocally connected (Vates et al., 1996). Among these sub-regions, L2a receives the strongest input from the core of Ov (nucleus ovoidalis), the primary auditory thalamus (Müller and Leppelsack, 1985; Rübsamen and Dörrscheidt, 1986; Hose et al., 1987). Secondary auditory areas — L2b, L3, and L1 — receive feed-forward input from L2a and thalamus, but also receive feedback from higher cortical areas such as CM. These hierarchically and reciprocally connected auditory areas are thought to be analogous to the early stages of mammalian auditory cortex, but the details of the homologies remain a subject of debate (Jarvis et al., 2005; Wang et al., 2010; Calabrese and Woolley, 2015).

Figure 1 with 3 supplements see all

Download asset Open asset

For zebra finches and other songbirds, temporal cues in song provide reliable information about the identity of the singer and are used for perceptual discrimination of songs (Gentner and Margoliash, 2003; Gentner et al., 2006; Grace et al., 2003; Shaevitz and Theunissen, 2007). The songbird auditory processing stream is well adapted to this information-processing task and reliably relays temporal information in conspecific song. In the zebra finch auditory system, there are neurons from midbrain to the highest levels of auditory association areas that respond with precise spike times to playback of conspecific song. This is true for both dense-spiking neurons and the highly selective, sparse-firing neurons recently described in the high-level auditory area, NCM (Schneider and Woolley, 2013), as well as in the auditory-motor association area, HVC (high vocal center) (Prather et al., 2008). Using a spectrotemporal receptive field (STRF) analysis, the effective temporal integration window of neurons in L2a, the first thalamorecipient zone, was observed to be very brief compared with responses one step further from the periphery in areas L1 and L3 (Kim and Doupe, 2011). The secondary areas (including L2b, L1, and L3) but not the primary zone (L2a) are recipients of significant feedback from high-order auditory areas (Vates et al., 1996). In combination, the results of these few studies suggest that an interesting transformation of temporal sequences could take place between primary and secondary zones in Field L.

Here, we developed new experimental paradigms to examine how temporally patterned auditory stimuli are transformed in the transition from the primary thalamorecipient zone, L2a, to the secondary auditory processing areas, L2b and L3. We first demonstrate that neurons responding to song in secondary areas, L2b and L3, become less synchronous in their relative response times yet more informative about the identity of specific syllables when compared to those in the primary area, L2a.

To zero in more closely on the nature of the transformation, we examined responses to a set of simplified auditory stimuli consisting of click sequences. The chosen stimuli were akin to ‘Morse code’ — the sounds differed only in the temporal ordering of intervals between clicks. These intervals were drawn from a distribution similar to the intervals between sub-syllabic acoustic transitions in zebra finch song (Gardner et al., 2001; Amador et al., 2013). For click sequence listening, a distinctive transformation of auditory responses was found between primary and secondary auditory zones. In the primary zone, each click elicited a similar low-latency response in all recorded neurons, and the structure of this response was largely insensitive to the temporal context of the click. One synapse further from the periphery, in secondary auditory areas, L2b and L3, neurons responded asynchronously and selectively, depending on the temporal context of the click. In effect, temporal sequences are transformed to distinct population vectors in the transition from primary to secondary auditory areas. In this process, temporal patterns come to be represented in a format that could directly form the basis of perceptual discriminations based on simple thresholds.

We next tested whether songbirds could discriminate different temporal click sequence patterns in an operant-training paradigm. A novel ‘restart-go’ operant paradigm, which we found effective for particularly challenging discrimination tests in zebra finches, was developed for this purpose. Using this training procedure, zebra finches rapidly learned to discriminate click sequences that were composed of song-like intervals. When the stimulus set was slowed by a factor of two, the strength of the temporal to spatial transformation in the secondary auditory was reduced, and there was a corresponding degradation of behavioral discrimination.

Taken together, these results indicate that the ascending auditory pathway in zebra finches transforms temporal sequences into distinct population vectors. This transformation applied to click sequences consisting of intervals that overlap with sub-syllabic acoustic structure in song, and may provide an important substrate for song perception and discrimination in sub-syllabic time-scales.

General note: the electrophysiological recordings reported here were gathered using four-shank, 32 channel silicon electrodes. From each bird, we recorded activity simultaneously from the primary thalamorecipient zone in the auditory area, Field L2a, and neighboring auditory areas in L2b and L3 (Figure 1a and b). All stimuli were presented in an interleaved fashion, and each animal was recorded acutely, with all data gathered in a single session. All data presented in figures and quantified below were gathered from well-sorted single-unit responses — a minority of recordings (Figure 3—figure supplement 3). The only exception to this rule is Figure 3—figure supplement 2, which includes a few channels of high SNR multi-unit traces that did not satisfy our criterion for single-unit isolation. These traces are marked with an asterisk. For additional details, see Methods.

Transformation of song responses in the auditory hierarchy

We first compared the temporal coding of song in primary (L2a) and secondary auditory areas (L2b and L3) of unanesthetized songbirds. Our intent was not to thoroughly catalog song responses, but rather to calibrate responses in order to design a set of synthetic stimuli that could be used for the remainder of the study. Primary and secondary recording sites were distinguished on the basis of the distinct response profiles found in the two areas (Figures 1a and 3a). This classification was confirmed by spatial mapping of the recording sites (Figure 1—figure supplement 1), showing that the primary cells were spatially segregated from the secondary neurons. Due to small anatomical and surgical variations and the small scale of the primary zone, this area could not be reliably identified by spatial coordinates alone.

Precise spike timing could be found in both primary and secondary areas in response to song. Focusing first on responses in the primary auditory area, L2a, we found a surprising degree of response synchrony across neurons and across birds (Figure 1a). The population peri-stimulus temporal histogram (PSTH) for each song was deeply modulated for neurons in L2a (Figure 1—figure supplement 3, Figure 2a is the histogram of inter-peak intervals in this population PSTH). By contrast, neurons in secondary auditory areas, L2b and L3, showed a broader repertoire of response profiles. This increase in the diversity of response timing leads to a decrease in the magnitude of the cross-correlation between the PSTHs of individual neurons in the secondary auditory areas relative to a similar cross-correlation performed in primary area, L2a (Figure 1d).

Figure 2

Download asset Open asset

Transformation of click-sequence responses in the auditory hierarchy

Our next objective was to examine whether a similar transformation from synchronous to asynchronous coding could be seen for more elementary stimuli consisting of irregularly spaced clicks. This synthetic stimulus would allow us to probe whether the sequence transformation from the primary to the secondary auditory areas requires complex spectral content. If secondary auditory neurons have more complex or more selective spectral receptive fields, the emergence of asynchronous coding in the secondary auditory areas could be explained on the basis of this acoustic selectivity alone. However, if the transformation from synchronous primary response to asynchronous secondary responses could be reproduced with click trains, the result would indicate that the auditory processing pathways contain intrinsic temporal dynamics that transform temporal sequences independent of spectral selectivity.

The chosen synthetic stimuli were three seconds long and composed of clicks separated by ten specific inter-click intervals (11, 14, 16, 20, 23, 26, 29, 34, 36 and 40 ms). We chose these intervals on the basis of the timescale of neural responses to birdsongs in L2a (Figure 2a and b). The inter-peak intervals of the population PSTH in response to these click sequences was similar to inter-peak intervals in response to natural song. In effect, we chose click patterns that, in the primary auditory area, elicited a temporal response that loosely overlapped with the natural song response. We note that the selected inter-click intervals are also similar to intervals between sub-syllabic acoustic transitions found in zebra finch song (Amador et al., 2013; Norton and Scharff, 2016). For comparison, Figure 2 also shows the L2a PSTH inter-peak interval histogram for click sequences slowed by a factor of two.

The duration of all ten click intervals summed together is 249 ms. The longer three-second sequences were built from 249 ms blocks, in which each block contains a permutation of the ten click intervals. In some stimuli, the blocks were repeating and in others non-repeating. For all sequences, the stimuli differed only in the ordering of click intervals. On timescales longer than the block duration, the statistical properties of all stimuli were equivalent. The set of stimuli used in this study can be seen in Figure 3—figure supplement 1. (Sample audio files are also provided. See supplementary file 1).

Raster plots for single units in primary and secondary auditory areas are shown in Figure 3a. (Example spike waveforms of single units and corresponding rasters are shown in Figure 3—figure supplement 3.) Raster plots for the full ensemble of single and multi-units are shown in Figure 3—figure supplement 2, including a breakdown of secondary cells into narrow (red) and broad-spiking (black) neuron waveforms. Only narrow units were found in the primary auditory area. (This figure is the only time in the paper that poorly sorted units, or ‘multi-units’ are included.) A distinct change in the temporal response to click sequences can be found in the transition from primary to secondary areas. In the primary auditory area, the click responses are fairly insensitive to the local context – to first approximation, each click evoked a synchronous, low latency response across channels, whereas secondary auditory areas were characterized by sparser and less synchronous responses that were more sensitive to the sequence context of the click (Figure 3—figure supplements 4 and 5). The click sequence, by definition, contains no significant spectral cues for frequencies above 100 Hz (the shortest interval in the click set was 11 ms, thus below the 100 Hz cutoff). Zebra finch hearing thresholds for pure tones are attenuated by about 20 dB relative to humans at 100 Hz (Okanoya and Dooling, 1987; Moore, 2007), and the fundamental frequency of conspecific song is typically 500 Hz or higher in zebra finches.

Figure 3 with 5 supplements see all

Download asset Open asset

As for song responses, the transition from primary to secondary thalamorecipient areas reveals a desynchronizing transformation that maps temporal click sequences onto distinct neuronal ensembles. For the click sequences used here, this transformation is even more apparent than for song responses. The diversification of neuronal responses increases the information about the preceding temporal context of a given click that the population vector contains. To demonstrate this, we computed phase-space trajectories of the population vector in response to click sequences, and then quantified the Euclidean distance between these phase-space trajectories. In this analysis, every neuron recorded defines a direction in a phasespace hypercube, and the average firing rate of the cell defines a position along the respective axis.

The phase-space trajectory for three cells in the primary auditory area and three cells in the secondary auditory areas during playback of two distinct sequences are shown in Figure 4a. In the primary auditory area, L2a, the phase-space trajectories of distinct stimuli overlap for all time points, meaning that the pattern of active cells contains little population-vector information that can distinguish the stimuli. By contrast, in secondary auditory areas, specific points in the phase-space trajectory diverge from one another in a stimulus-dependent manner. That is, the pattern of cell responses in secondary auditory areas contains information about one or more intervals preceding the click. To summarize simply – there are particular configurations of active cells that occur only during playback of one stimulus or another — a useful feature for a system that is tuned to make fine discriminations about temporal sequence patterns.

Figure 4 with 1 supplement see all

Download asset Open asset

Figure 4—source data 1

Source data for ROC curve.

This zip file contains spike-timing data used for the ROC analysis shown in Figure 4b. Spike times of 10 different cells recorded in primary or secondary auditory areas are included in folders with corresponding names. For simple visualization of spike rasters, Matlab source code (DataLoad.m) is also provided.

https://doi.org/10.7554/eLife.18205.014

Download elife-18205-fig4-data1-v2.zip

To quantify the degree to which the click stimuli can be distinguished on the basis of the neural responses, we defined a simple decoding mechanism based on the population vector of the ensemble response (see Methods for details). In this decoding, the discriminability of the sequence at a particular time is given by the distance in phase space to the nearest trajectory belonging to a different stimulus. The power of this ‘spatial’ code for sequence discrimination is quantified through an ROC (receiver operating characteristic) analysis in Figure 4b. We analyzed coding in primary and secondary areas using the ROC analysis, using a fixed number of single unit recordings (n = 10) in both cases. In the secondary auditory areas, but not the primary thalamorecipient area, temporal sequences are mapped onto distinct population patterns, revealing a better sequence decoding in the ROC analysis (spike times of the units used in this analysis are also provided in Figure 4—source data 1). We repeated this analysis for just the first 500 ms of the stimulus, and still found a high degree of sequence discriminability in the secondary auditory areas (Figure 4—figure supplement 1). This shorter analysis is more directly relevant to the behavioral discriminations reported below, as trained birds performing behavioral discriminations typically respond within this time frame (Figure 6—figure supplement 2). To further validate this approach, we applied the same analysis to the PSTH of the song syllable responses (n = 13 syllables, Figure 1a) and found an increase in syllable discriminability in the secondary auditory area (Figure 1—figure supplement 2). Given the rich spectral content of song relative to clicks, the primary area, L2a, already shows a high degree of response selectivity, better than that in the response to the click sequences.

We next repeated the click electrophysiology using a stimulus set composed of intervals twice as long as those in the first stimulus set (22–80 ms, rather than 11–40 ms, Figure 2c). This change in stimulus timescale had a minimal impact on spike rate in the secondary auditory cortex (Figure 5—figure supplement 1), but resulted in a significant change in the power of the temporal to spatial transformation. Using the same phase-plane ROC analysis, we found that the timescale dilation led to reduced sequence discrimination in secondary auditory areas (Figure 5).

Figure 5 with 1 supplement see all

Download asset Open asset

Behavioral recognition of click sequences

The preceding electrophysiology experiments demonstrated a transformation of click responses to distinct population vectors in the secondary auditory areas of unanesthetized zebra finches. As a result, areas downstream of secondary auditory areas could, in principle, solve a click-sequence classification problem by applying a simple summation and threshold to subsets of secondary cell inputs. Given the robust transformation of temporal click sequences in zebra finch auditory areas, we next sought to determine whether songbirds could be trained to behaviorally discriminate this class of artificial stimuli, and whether or not properties of the electrophysiological responses correlated with behavioral discriminations.

Songbirds were trained using a new operant-training procedure developed for this study (Figure 6—figure supplement 1). We call this automated training ‘reset-go'. A detailed description of the training procedure can be found in the Methods. In essence, a bird can demonstrate learning through two behaviors — by interrupting the playback of a non-rewarding stimulus to ‘request’ the reset of an unfavorable trial or by accessing the water port during playback of rewarding stimuli. In all experiments, two sounds were presented — a rewarded stimulus (click sequence 2 from Figure 3—figure supplement 1) and non-rewarded stimulus (click sequence 1 or 9 from Figure 3—figure supplement 1). Zebra finches in this task were mildly water restricted, and worked for 1–5 μl drops of water, routinely performing a thousand trials in a five-hour training session.

Figure 6a reveals the time-course of discrimination learning for one bird. Ten days after the initiation of training, this bird would interrupt the playback of the unrewarded stimulus (sequence 1) within three seconds and access the water port while the rewarded stimulus (sequence 2) was presented. Figure 6b shows summary statistics for learning in eight birds trained to discriminate sequence 1 vs. sequence 2. Figure 6—source data 1 documents the groups of birds trained and Figure 6—figure supplement 2 shows the time-course of learning for the various groups. The detailed training procedure is described in the Methods. Within a population of trained birds (n = 53), a majority (n = 35 birds) showed high levels of performance (d’ > 1) within 14 days of training onset, revealing that songbirds could readily learn to discriminate the fast temporal click sequences used in this study.

Figure 6 with 2 supplements see all

Download asset Open asset

Figure 6—source data 1

Summary of training.

The success of operant training was determined on the basis of the d-prime score. When d’ is greater than 1, the bird was deemed successful in learning the task. In this table, the number of birds that succeeded in operant training for click sequence discrimination (d’ > 1) out of the total number of birds is shown. For example, 8 out of 10 birds succeeded in two-stage training to discriminate sequence 9 and 2.

https://doi.org/10.7554/eLife.18205.019

Download elife-18205-fig6-data1-v2.docx

Catch trials probe the nature of auditory discrimination

To probe the underlying nature of the auditory discrimination, we examined catch trials for two conditions. For time-reversed click stimuli, behavior fell to chance levels (Figure 7a), indicating that the ordering of the click intervals was critical to the behavioral discrimination. The next test examined cyclic permutations of the training stimuli. Rather than beginning playback at the normal starting interval of each sequence, the cyclic permutation initiated each stimulus at a random click interval in the three second stimulus – effectively a phase shift in the stimulus. For this group of catch trials, a small decrease in performance was found when the cyclic stimulus was first introduced, but within four days, performance returned to baseline (Figure 7b). The conclusion from this is that the birds are listening for patterned sequences of intervals irrespective of their absolute time of occurrence relative to the onset of the trial.

Figure 7

Download asset Open asset

Songbirds form detailed auditory memories for complex songs, and these memories serve to guide imitative vocal learning (Nottebohm, 1972; Brainard and Doupe, 2002; Bolhuis and Gahr, 2006; Gardner et al., 2005). In parallel, a range of songbird species can perform at high levels in operant tasks involving song and synthetic stimulus discrimination (Gess et al., 2011; Sturdy and Weisman, 2006; Cynx and Nottebohm, 1992; Scharff et al., 1998; Stripling et al., 2003; Abe and Watanabe, 2011). While songbird auditory performance has been well documented, the network mechanisms underlying song discriminations have not been studied. In particular, one of the least understood aspects of auditory sequence processing concerns the transformations applied to complex temporal sequences (Mauk and Buonomano, 2004).

The present study provides insight into the processing of a simple class of temporal sequences composed of irregularly spaced clicks. We find that after the stimulus passes through the primary thalamorecipient zone — L2a, L2b, and L3 — these temporal sequences are transformed into distinct population vectors or ‘spatial codes'. The mapping of temporal patterns to spatial patterns or ensemble codes provides an opportunity for downstream neurons to perform stimulus discrimination using simple linear classifiers to act on the population vector. For the click stimuli used in this study, reliable discriminations could be made on the basis of the distinct population vectors that arise in L2b and L3, binned in 5 ms time bins.

Operant training revealed that songbirds readily learn to discriminate the Morse-code like click stimuli. The fast-click sequences were behaviorally discriminable with high accuracy for a majority of trained birds. Surprisingly, no animals learned to discriminate click sequences that were slowed by a factor of two, even though secondary auditory areas responded with similar spike rates to the slower stimulus. The slowed sequences evoked inter-peak intervals in primary auditory area PSTH that were longer than the typical intervals between peaks in the PSTH during natural song exposure. We suggest that the ascending auditory pathway in the transition from L2a to L2b and L3 is tuned to process temporal events on the faster timescale (11–40 ms) in a manner that is particularly useful for song memorization and discrimination.

We mention two caveats in the present study. First, the high-pass cutoff frequency of the loudspeakers was 3 kHz. (High frequency tweeters were used for stimulus delivery limiting the spectral content of each click.) We do not know how the spectral content of the click impacts the behavioral discrimination of the slower sequences. In another prior study, zebra finches were able to discriminate sequences of beeps spaced by intervals of up to 300 ms — intervals much longer than those used in our study (van der Aa et al., 2015). It is likely that brief clicks and longer tones tap into auditory processing pathways with distinct temporal dynamics, explaining the performance difference. In addition, many details of the temporal discrimination tasks were different in the two studies, and the distinct results may also relate to these task differences. Additional tests will be needed to determine whether or not the spectral content of each click impacts the behavioral performance. Opportunities also exist to further examine the ability of the zebra finch to generalize temporal pattern recognition through time-dilations (Nagel et al., 2010).

The second caveat is that the single-unit ensembles studied here were ‘virtual ensembles’ recorded in different animals; noise correlations within animals could further impact discrimination in ways that were not addressed here (Zohary et al., 1994; Abbott and Dayan, 1999). While we did not acquire enough high-quality single-unit data to perform the ROC analysis for individual animals, enough units were recorded simultaneously to reveal the transformation qualitatively from primary to secondary responses in summary raster plots (Figure 3—figure supplements 2 and 3). These rasters support the view that the sequence transformation described for virtual ensembles will also hold for ensembles of neurons in individual birds.

Much theoretical interest has focused on the question of how brains composed of neurons with short intrinsic timescales can process long-timescale stimuli and generate long-timescale behaviors (Lashely, 2004). For temporal stimuli composed of identical units such as clicks, intrinsic cellular or circuit mechanisms must bridge intervals of time from one interval to the next in order to create sequence-specific population responses. To encode the history of the stimulus in the present state of the network, synfire chains, avalanches, or more complex transient dynamics in recurrent networks have all been proposed (Abeles, 1991; Grossberg, 1969; Maass et al., 2002). In other models, persistent currents in single cells bridge intervals of time (Egorov et al., 2002). In each of these models, intrinsic dynamics of cortical cells or circuits are used to transfer information about past events into the network responses at a given time.

One effective way of transferring information about prior events into current responses is through feedback connections. The primary auditory area (L2a) in songbirds reportedly receives no feedback from higher-level auditory zones (Vates et al., 1996), and the synchronous, low-latency responses in this region may reflect a feed-forward response to thalamic drive. By contrast, all other areas, including the secondary auditory zones examined here (L2b and L3), are more densely interconnected both with each other and with higher-level auditory areas. This anatomical distinction raises the possibility that L2b and L3, but not the primary auditory area, L2a, can sustain reverberant activity that could underlie the temporal sequence transformation observed in L2b and L3. Relevant theoretical constructs for this model include liquid state machine theories (Maass et al., 2002). By way of illustration, Figure 8 reveals the output of a simple reverberant model that recapitulates key features of the observed dynamics. In this case, the model is simply a linear dynamical system driven by click sequences and additive noise, and tick marks represent time points when the vector $\left(v\right)$ crosses arbitrary threshold amplitude.

Figure 8

Download asset Open asset

In this example, matrix $M$ is a random anti-symmetric matrix, with all imaginary eigenvalues, and $\eta$ is a random noise term. By choosing the time-constants $\alpha$ and $\gamma$ appropriately, the model can produce patterns that resemble spike rasters observed in L2b and L3. Figure 8, generated by this linear system, simply illustrates the point that even the simplest recurrent dynamical systems have the capability to transform click-sequence information into distinct population vectors when properly tuned. In this example, the anti-symmetric matrix, $M$, provides a form of ‘critical tuning’ in which multiple oscillatory timescales are equally excitable, providing richer temporal dynamics than would occur for a typical nonsymmetric random connectivity matrix (Magnasco et al., 2009). While the hypothesis that recurrence explains the auditory sequence transformation is appealing, new experimental studies are needed to examine the role of recurrent dynamics in temporal sequence processing in the auditory pallium of songbirds.

While the reverberant models provide an attractive explanation for the sequence transformation observed in L2b and L3, the behavioral discriminations that the birds exhibited here cannot be taken as evidence that supports the reverberant model, strictly speaking. A purely feed-forward counter-hypothesis is that the neurons in secondary auditory areas could demonstrate biophysical integration timescales that solve the sequence discrimination through single-cell properties. To illustrate this hypothesis, we first smoothed the click sequences used for behavior training with three different rectangular windows of timescale $T$ or shorter and built phase-plane traces of these hypothetical ‘smoothing units’ in 3D space. In this case, three different smoothing windows correspond to hypothetical units with different integration timescales. We then analyzed the minimal time-scale $T$ for which the behaviorally trained sequences could be perfectly segregated in the ROC analysis performed earlier for actual neural sequences. From this analysis, we found that phase plane traces used in the behavioral studies can be perfectly separated if the width of the longest rectangular window was 100 ms or greater. To state this more simply, while the sequences presented to the birds were 3 s long, click rates measured in time bins as short as 100 ms provide a population vector that is adequate for sequence discrimination. This 100 ms timescale cannot rule out either feed-forward single-cell biophysics or recurrent dynamics as contributors to the sequence transformation.

While the circuit mechanisms remain to be established, this study serves to demonstrate both a distinctive transformation of temporal sequences in the transition from L2a to higher-order areas, L2b and L3, and a behavioral capacity of zebra finches to discriminate synthetic click sequences. The transformation of temporal sequences to distinct population vectors may underlie the songbird’s advanced discrimination abilities for temporally structured conspecific song.

All procedures were approved by the Institutional Animal Care and Use Committee of Boston University.

1. 1. Abbott LF
   2. Dayan P

   (1999) The effect of correlated variability on the accuracy of a population code

   Neural Computation 11:91–101.

   https://doi.org/10.1162/089976699300016827

   • PubMed
   • Google Scholar
2. 1. Abe K
   2. Watanabe D

   (2011) Songbirds possess the spontaneous ability to discriminate syntactic rules

   Nature Neuroscience 14:1067–1074.

   https://doi.org/10.1038/nn.2869

   • PubMed
   • Google Scholar
3. Book

   1. Abeles M

   (1991) Corticonics

   Cambridge University Press.

   https://doi.org/10.1017/CBO9780511574566

   • Google Scholar
4. 1. Amador A
   2. Perl YS
   3. Mindlin GB
   4. Margoliash D

   (2013) Elemental gesture dynamics are encoded by song premotor cortical neurons

   Nature 495:59–64.

   https://doi.org/10.1038/nature11967

   • PubMed
   • Google Scholar
5. 1. Bolhuis JJ
   2. Gahr M

   (2006) Neural mechanisms of birdsong memory

   Nature Reviews. Neuroscience 7:347–357.

   https://doi.org/10.1038/nrn1904

   • PubMed
   • Google Scholar
6. 1. Brainard MS
   2. Doupe AJ

   (2002) What songbirds teach us about learning

   Nature 417:351–358.

   https://doi.org/10.1038/417351a

   • PubMed
   • Google Scholar
7. 1. Cade TJ
   2. Tobin CA
   3. Gold A

   (1965) Water economy and metabolism of two estrildine finches

   Physiological Zoology 38:9–33.

   https://doi.org/10.1086/physzool.38.1.30152342

   • Google Scholar
8. 1. Calabrese A
   2. Woolley SM

   (2015) Coding principles of the canonical cortical microcircuit in the avian brain

   PNAS 112:3517–3522.

   https://doi.org/10.1073/pnas.1408545112

   • PubMed
   • Google Scholar
9. 1. Cynx J
   2. Nottebohm F

   (1992) Testosterone facilitates some conspecific song discriminations in castrated zebra finches (Taeniopygia guttata)

   PNAS 89:1376–1378.

   https://doi.org/10.1073/pnas.89.4.1376

   • PubMed
   • Google Scholar
10. 1. Egorov AV
    2. Hamam BN
    3. Fransén E
    4. Hasselmo ME
    5. Alonso AA

    (2002) Graded persistent activity in entorhinal cortex neurons

    Nature 420:173–178.

    https://doi.org/10.1038/nature01171

    • PubMed
    • Google Scholar
11. 1. Fortune ES
    2. Margoliash D

    (1992) Cytoarchitectonic organization and morphology of cells of the field L complex in male zebra finches (Taenopygia guttata)

    The Journal of Comparative Neurology 325:388–404.

    https://doi.org/10.1002/cne.903250306

    • PubMed
    • Google Scholar
12. 1. Gardner T
    2. Cecchi G
    3. Magnasco M
    4. Laje R
    5. Mindlin GB

    (2001) Simple motor gestures for birdsongs

    Physical Review Letters 87:208101.

    https://doi.org/10.1103/PhysRevLett.87.208101

    • PubMed
    • Google Scholar
13. 1. Gardner TJ
    2. Naef F
    3. Nottebohm F

    (2005) Freedom and rules: the acquisition and reprogramming of a bird's learned song

    Science 308:1046–1049.

    https://doi.org/10.1126/science.1108214

    • PubMed
    • Google Scholar
14. 1. Gentner TQ
    2. Fenn KM
    3. Margoliash D
    4. Nusbaum HC

    (2006) Recursive syntactic pattern learning by songbirds

    Nature 440:1204–1207.

    https://doi.org/10.1038/nature04675

    • PubMed
    • Google Scholar
15. 1. Gentner TQ
    2. Margoliash D

    (2003) Neuronal populations and single cells representing learned auditory objects

    Nature 424:669–674.

    https://doi.org/10.1038/nature01731

    • PubMed
    • Google Scholar
16. 1. Gess A
    2. Schneider DM
    3. Vyas A
    4. Woolley SM

    (2011) Automated auditory recognition training and testing

    Animal Behaviour 82:285–293.

    https://doi.org/10.1016/j.anbehav.2011.05.003

    • PubMed
    • Google Scholar
17. 1. Grace JA
    2. Amin N
    3. Singh NC
    4. Theunissen FE

    (2003) Selectivity for conspecific song in the zebra finch auditory forebrain

    Journal of Neurophysiology 89:472–487.

    https://doi.org/10.1152/jn.00088.2002

    • PubMed
    • Google Scholar
18. Book

    1. Green DM
    2. Swets JA

    (1966)

    Signal Detection Theory and Psychophysics

    New York: Wiley.

    • Google Scholar
19. 1. Grossberg S

    (1969) On learning of spatiotemporal patterns by networks with ordered sensory and motor components 1. Excitatory components of the cerebellum

    Studies in Applied Mathematics 48:105–132.

    https://doi.org/10.1002/sapm1969482105

    • Google Scholar
20. 1. Hose B
    2. Langner G
    3. Scheich H

    (1987) Topographic representation of periodicities in the forebrain of the mynah bird: one map for pitch and rhythm?

    Brain Research 422:367–373.

    https://doi.org/10.1016/0006-8993(87)90946-2

    • PubMed
    • Google Scholar
21. 1. Jarvis ED
    2. Güntürkün O
    3. Bruce L
    4. Csillag A
    5. Karten H
    6. Kuenzel W
    7. Medina L
    8. Paxinos G
    9. Perkel DJ
    10. Shimizu T
    11. Striedter G
    12. Wild JM
    13. Ball GF
    14. Dugas-Ford J
    15. Durand SE
    16. Hough GE
    17. Husband S
    18. Kubikova L
    19. Lee DW
    20. Mello CV
    21. Powers A
    22. Siang C
    23. Smulders TV
    24. Wada K
    25. White SA
    26. Yamamoto K
    27. Yu J
    28. Reiner A
    29. Butler AB
    30. Avian Brain Nomenclature Consortium

    (2005) Avian brains and a new understanding of vertebrate brain evolution

    Nature Reviews. Neuroscience 6:151–159.

    https://doi.org/10.1038/nrn1606

    • PubMed
    • Google Scholar
22. 1. Kim G
    2. Doupe A

    (2011) Organized representation of spectrotemporal features in songbird auditory forebrain

    Journal of Neuroscience 31:16977–16990.

    https://doi.org/10.1523/JNEUROSCI.2003-11.2011

    • PubMed
    • Google Scholar
23. 1. Lashely K

    (2004)

    First Language Acquisition: The Essential Readings

    316–334, The probelm of serial order in behavior, First Language Acquisition: The Essential Readings.

    • Google Scholar
24. 1. Ludwig KA
    2. Miriani RM
    3. Langhals NB
    4. Joseph MD
    5. Anderson DJ
    6. Kipke DR

    (2009) Using a common average reference to improve cortical neuron recordings from microelectrode arrays

    Journal of Neurophysiology 101:1679–1689.

    https://doi.org/10.1152/jn.90989.2008

    • PubMed
    • Google Scholar
25. 1. Maass W
    2. Natschläger T
    3. Markram H

    (2002) Real-time computing without stable states: a new framework for neural computation based on perturbations

    Neural Computation 14:2531–2560.

    https://doi.org/10.1162/089976602760407955

    • PubMed
    • Google Scholar
26. 1. Magnasco MO
    2. Piro O
    3. Cecchi GA

    (2009) Self-tuned critical anti-Hebbian networks

    Physical Review Letters 102:258102.

    https://doi.org/10.1103/PhysRevLett.102.258102

    • PubMed
    • Google Scholar
27. 1. Mandelblat-Cerf Y
    2. Las L
    3. Denissenko N
    4. Fee M

    (2013) A role for descending auditory cortical projections in songbird vocal learning

    eLife 3:e02152.

    https://doi.org/10.7554/eLife.02152

    • Google Scholar
28. 1. Mauk MD
    2. Buonomano DV

    (2004) The neural basis of temporal processing

    Annual Review of Neuroscience 27:307–340.

    https://doi.org/10.1146/annurev.neuro.27.070203.144247

    • PubMed
    • Google Scholar
29. Book

    1. Moore BCJ

    (2007)

    An Introduction to the Psychology of Hearing (5 ed)

    Academic Press.

    • Google Scholar
30. 1. Müller CM
    2. Leppelsack HJ

    (1985) Feature extraction and tonotopic organization in the avian auditory forebrain

    Experimental Brain Research 59:587–599.

    https://doi.org/10.1007/BF00261351

    • PubMed
    • Google Scholar
31. 1. Nagel KI
    2. McLendon HM
    3. Doupe AJ

    (2010) Differential influence of frequency, timing, and intensity cues in a complex acoustic categorization task

    Journal of Neurophysiology 104:1426–1437.

    https://doi.org/10.1152/jn.00028.2010

    • PubMed
    • Google Scholar
32. 1. Norton P
    2. Scharff C

    (2016) "Bird Song Metronomics": Isochronous Organization of Zebra Finch Song Rhythm

    Frontiers in Neuroscience 10:309.

    https://doi.org/10.3389/fnins.2016.00309

    • PubMed
    • Google Scholar
33. 1. Nottebohm F

    (1972) The origins of vocal learning

    The American Naturalist 106:116–140.

    https://doi.org/10.1086/282756

    • Google Scholar
34. 1. Okanoya K
    2. Dooling RJ

    (1987) Hearing in passerine and psittacine birds: a comparative study of absolute and masked auditory thresholds

    Journal of Comparative Psychology 101:7–15.

    https://doi.org/10.1037/0735-7036.101.1.7

    • PubMed
    • Google Scholar
35. 1. Pham DT
    2. Dimov SS
    3. Nguyen CD

    (2005) Selection of K in K-means clustering

    Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 219:103–119.

    https://doi.org/10.1243/095440605X8298

    • Google Scholar
36. 1. Picardo MA
    2. Katlowitz KA
    3. Okobi DE
    4. Benezra SE
    5. Clary RC
    6. Merel J
    7. Paninski L
    8. Long MA

    (2015)

    Neuroscience Meeting Planner

    Analyzing the population dynamics underlying a complex motor act. Program No. 181.01. 2015, Neuroscience Meeting Planner, Chicago, IL, Society for Neuroscience, Online.

    • Google Scholar
37. 1. Prather JF
    2. Peters S
    3. Nowicki S
    4. Mooney R

    (2008) Precise auditory-vocal mirroring in neurons for learned vocal communication

    Nature 451:305–310.

    https://doi.org/10.1038/nature06492

    • PubMed
    • Google Scholar
38. 1. Quiroga RQ
    2. Nadasdy Z
    3. Ben-Shaul Y

    (2004) Unsupervised spike detection and sorting with wavelets and superparamagnetic clustering

    Neural Computation 16:1661–1687.

    https://doi.org/10.1162/089976604774201631

    • PubMed
    • Google Scholar
39. 1. Rübsamen R
    2. Dörrscheidt GJ

    (1986) Tonotopic organization of the auditory forebrain in a songbird, the European starling

    Journal of Comparative Physiology A 158:639–646.

    https://doi.org/10.1007/BF00603820

    • Google Scholar
40. 1. Scharff C
    2. Nottebohm F
    3. Cynx J

    (1998) Conspecific and heterospecific song discrimination in male zebra finches with lesions in the anterior forebrain pathway

    Journal of Neurobiology 36:81–90.

    https://doi.org/10.1002/(SICI)1097-4695(199807)36:1<81::AID-NEU7>3.0.CO;2-6

    • PubMed
    • Google Scholar
41. 1. Schneider DM
    2. Woolley SM

    (2013) Sparse and background-invariant coding of vocalizations in auditory scenes

    Neuron 79:141–152.

    https://doi.org/10.1016/j.neuron.2013.04.038

    • PubMed
    • Google Scholar
42. 1. Shaevitz SS
    2. Theunissen FE

    (2007) Functional connectivity between auditory areas field L and CLM and song system nucleus HVC in anesthetized zebra finches

    Journal of Neurophysiology 98:2747–2764.

    https://doi.org/10.1152/jn.00294.2007

    • PubMed
    • Google Scholar
43. 1. Stripling R
    2. Milewski L
    3. Kruse AA
    4. Clayton DF

    (2003) Rapidly learned song-discrimination without behavioral reinforcement in adult male zebra finches (Taeniopygia guttata)

    Neurobiology of Learning and Memory 79:41–50.

    https://doi.org/10.1016/S1074-7427(02)00005-9

    • PubMed
    • Google Scholar
44. 1. Sturdy CB
    2. Weisman RG

    (2006) Rationale and methodology for testing auditory cognition in songbirds

    Behavioural Processes 72:265–272.

    https://doi.org/10.1016/j.beproc.2006.03.007

    • PubMed
    • Google Scholar
45. 1. van der Aa J
    2. Honing H
    3. ten Cate C

    (2015) The perception of regularity in an isochronous stimulus in zebra finches (Taeniopygia guttata) and humans

    Behavioural Processes 115:37–45.

    https://doi.org/10.1016/j.beproc.2015.02.018

    • PubMed
    • Google Scholar
46. 1. Vates GE
    2. Broome BM
    3. Mello CV
    4. Nottebohm F

    (1996) Auditory pathways of caudal telencephalon and their relation to the song system of adult male zebra finches

    The Journal of Comparative Neurology 366:613–642.

    https://doi.org/10.1002/(SICI)1096-9861(19960318)366:4&amp;lt;613::AID-CNE5&amp;gt;3.0.CO;2-7

    • PubMed
    • Google Scholar
47. 1. Wang Y
    2. Brzozowska-Prechtl A
    3. Karten HJ

    (2010) Laminar and columnar auditory cortex in avian brain

    PNAS 107:12676–12681.

    https://doi.org/10.1073/pnas.1006645107

    • PubMed
    • Google Scholar
48. 1. Zohary E
    2. Shadlen MN
    3. Newsome WT

    (1994) Correlated neuronal discharge rate and its implications for psychophysical performance

    Nature 370:140–143.

    https://doi.org/10.1038/370140a0

    • PubMed
    • Google Scholar

Author details

1. Yoonseob Lim

   1. Department of Cognitive and Neural Systems, Boston University, Boston, United States
   2. Convergence Research Center for Diagnosis, Treatment, and Care System for Dementia, Korea Institute of Science and Technology, Seoul, Korea

   Contribution

   YL, Designed the study, Collected data, Wrote the paper and analyzed the data

   Competing interests

   No competing interests declared.

2. Ryan Lagoy

   Department of Electrical and Computer Engineering, Boston University, Boston, United States

   Contribution

   RL, Collected data

   Competing interests

   No competing interests declared.

3. Barbara G Shinn-Cunningham

   Department of Biomedical Engineering, Boston University, Boston, United States

   Contribution

   BGS-C, Designed the study

   Competing interests

   BGS-C: Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0002-5096-5914

4. Timothy J Gardner

   1. Department of Biomedical Engineering, Boston University, Boston, United States
   2. Department of Biology, Boston University, Boston, United States

   Contribution

   TJG, Designed the study, Wrote the paper

   For correspondence

   timothyg@bu.edu

   Competing interests

   No competing interests declared.

   "This ORCID iD identifies the author of this article:" 0000-0002-1744-3970

• 1,837

  views

• 377

  downloads

• 20

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.