Cool songs

Article metrics

Cooling a specific cluster of neurons in songbirds' brains slows song tempo without changing other acoustic features. This clever technique could be used for understanding neural control of other complex behaviours.

Complex behaviours, ranging from speech and typing to dancing and swimming, all require careful coordination of dozens and often hundreds of different muscles. The resulting behaviour frequently shows hierarchical structure, with sequences composed of primitive movements, each of which is composed of even more basic movements. How is all this coordinated? Researchers have long postulated that the timing of behavioural components is encoded in specialized brain circuits. Without such circuits, movement would quickly become disorganized, like an orchestra performing without a conductor. On page 189 of this issue, Long and Fee1 provide a remarkable example of such specialization in the brain of the zebra finch by using an equally remarkable experimental approach.

The zebra finch's song consists of a rich set of acoustic features produced with very precise timing. Yet localizing the brain regions that coordinate these features has been tricky. A traditional approach is to measure how neural lesions or electrical stimulation alter song output. But with complex behaviours such as birdsong, such perturbations will often initiate a cascade of disturbance that causes the action to fall apart, making it impossible to isolate the function of the brain area that was initially perturbed.

To circumvent this problem, Long and Fee put a new twist on an old technique. Localized cooling has long been used as a method to selectively and reversibly inactivate particular brain areas2. Long and Fee, however, applied only slight cooling, focusing on nucleus HVC — an area of the bird's brain crucial to song production. They find that cooling HVC in this way stretches the temporal fabric of the song, slowing tempo but leaving acoustic features such as pitch largely unchanged. Moreover, the degree of cooling determined the amount of slowing, allowing the authors to effectively 'dial in' a desired tempo.

The results challenge previous notions about how HVC contributes to the hierarchical arrangement of song acoustics. Songs are organized into motifs, consisting of stereotyped sequences of syllables; syllables in turn are composed of one or more vocalizations with distinct acoustic features called notes. Early studies had suggested3,4 that this organization mapped directly onto different brain areas, with nucleus HVC driving syllable sequence and timing, and RA (the robust nucleus of the arcopallium), a region directly downstream of HVC, determining how each syllable is produced. Long and Fee reasoned that if this were the case, cooling HVC would only stretch the times between syllable onsets, lengthening silent gaps, and would leave syllable lengths unchanged (Fig. 1a,b). However, they find that cooling HVC stretches syllables and gaps alike (Fig. 1c), whereas cooling RA does not affect the song tempo. It therefore seems that nucleus HVC plays a central part in controlling song timing at all levels of the song hierarchy.

Figure 1: Effects of cooling on song.

a, A normal song can be divided into intermittent syllables separated by inter-syllable gaps. b, It was thought that the song nucleus HVC controls only the onset times of syllables, and so its cooling would lengthen the inter-syllable gaps without affecting syllable lengths. c, Long and Fee1 show that, in fact, HVC influences timing throughout the song as its cooling stretches both syllables and the gaps between them.

The authors also find that the lengthening of any given song segment in response to cooling is proportional to its original length. Studies of motor control in humans have shown a similar scaling of change in timing with the duration of movement, and several researchers have proposed that this scaling property is a signature of motor-planning circuits5,6. Previous recordings in HVC suggested7 that song is driven by the serial activation of distinct sets of neurons every 5–10 milliseconds. The most likely explanation is that these neuronal populations are chained together in series, and together constitute a 'motor tape' that is read by RA to produce specific combinations of song features. Slow down the tape and you slacken the song tempo. The longer the song segment, the more motor tape it uses and so the longer it will become on cooling.

But there is a major complication with the hypothesis that timing is localized in HVC: there are actually two HVCs, one in each hemisphere of the bird's forebrain. So are there really two motor tapes producing song? What if each has a slightly different tempo during a song rendition, just by chance? Long and Fee examined this by cooling each HVC separately, and found that, indeed, both contribute to tempo, with the left HVC being dominant during some parts of the song, whereas the right HVC prevails during others.

What keeps the two HVCs from drifting completely out of alignment? There are no direct connections between the two halves of the brain at the level of HVC and RA, so synchronizing signals must travel through bilateral circuitry in the brainstem and thalamus that is downstream of RA (Fig. 2). This information can then be fed back up to both HVCs. Intriguingly, similar circuits that loop through the brainstem are known8 to be instrumental in coordinating other complex behaviours, such as learned eye-movement sequences in primates. But figuring out how such circuits work can be mind-bending: during the song, the input of signals to a given nucleus — whether HVC, RA or those within the brainstem — can be influenced by activities anywhere else within the circuit, including an earlier output of the same nucleus itself. How can one particular area initiate timing?

Figure 2: Song circuitry.

As shown in this simplified diagram, the feedforward pathway mediating singing in birds consists of inputs from nucleus HVC to nucleus RA to motor neurons in the brainstem nuclei. RA also sends feedback signals to HVC by way of the brainstem and thalamus. Long and Fee1 show that cooling either left or right HVC can slow song, whereas cooling RA alone has no influence. As there are no direct connections between the left and right hemispheres at the level of HVC or RA, feedback signals passing through the brainstem or thalamus have an essential role in synchronizing the two sides of the song motor circuit.

Simultaneous recordings in both HVCs of singing birds have indicated9 that synchronization may occur at discrete points in a song, such as syllable onsets. Drawing on this finding, Long and Fee propose that synchronizing signals originate in both HVCs and rapidly cycle through the brainstem, crossing over to the other side, to ensure that each HVC continues on the same beat. But it is also possible that the brainstem itself initiates synchronization. If so, cooling particular brainstem areas might also slow tempo, albeit with different patterns of stretching — such as that shown in Figure 1b. Other studies have suggested10 that the influence of brainstem feedback signals is not restricted to specific times during the song. So another possibility is that brainstem feedback acts throughout the song to help keep the two HVCs in step.

The zebra finch has provided a wonderful model for studying a temporally complex behaviour. By demonstrating that nucleus HVC can influence song tempo at all levels of song organization, Long and Fee have taken a crucial step towards deciphering the motor code for song production. Although the findings leave open many questions about the coordination of neural signals during song production, the cooling method itself points in an exciting direction for future studies: by slowing down the neural processing in one brain area, simultaneous electrophysiological recordings from other areas can be used to track how distortion of timing propagates through the system. Given that mild cooling has been shown to slow down neural processing in various systems11, 'cool and slow' has tremendous potential for understanding the temporal coordination of brain activity.


  1. 1

    Long, M. A. & Fee, M. A. Nature 456, 189–194 (2008).

  2. 2

    Mark, V. H., Chato, J. C., Eastman, F. G., Aronow, S. & Ervin, F. R. Science 134, 1520–1521 (1961).

  3. 3

    Vu, E. T., Mazurek, M. E. & Kuo, Y. J. Neurosci. 14, 6924–6934 (1994).

  4. 4

    Yu, A. C. & Margoliash, D. Science 273, 1871–1875 (1996).

  5. 5

    Wing, A. M. & Kristofferson, A. B. Percept. Psychophys. 14, 5–12 (1973).

  6. 6

    Ivry, R. B. & Richardson, T. C. Brain Cogn. 48, 117–132 (2002).

  7. 7

    Hahnloser, R. H. R., Kozhevnikov, A. A. & Fee, M. S. Nature 419, 65–70 (2002).

  8. 8

    Sommer, M. A. & Wurtz, R. H. Science 296, 1480–1482 (2002).

  9. 9

    Schmidt, M. F. J. Neurophysiol. 90, 3931–3949 (2003).

  10. 10

    Ashmore, R. C., Wild, J. M. & Schmidt, M. F. J. Neurosci. 25, 8543–8554 (2005).

  11. 11

    Janssen, R. Neurosci. Biobehav. Rev. 16, 399–413 (1992).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Glaze, C., Troyer, T. Cool songs. Nature 456, 187–188 (2008) doi:10.1038/456187a

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.