Understanding how early auditory memories are laid down could help to explain their role in vocal development. Some ingenious experiments in birds provide fresh ideas about how such memories are represented.
A central discovery that emerged from the early studies of Konrad Lorenz and Nikolaas Tinbergen was that animals have innate predispositions that guide the learning of species-appropriate signals. Birdsongs are a good example of such signals. Young songbirds must learn their specific songs by listening to adults, and research has provided insight into how auditory memories are formed under species-specific constraints, and how those memories help to guide song development. Yet the organization of the ‘acquired template’ for song learning that is formed from the early experience of hearing other birds sing remains a mystery, at both the neurobiological and (more surprisingly) the behavioural levels. Elsewhere in this issue (page 753), Rose et al.1 describe certain features of that template: they show how white-crowned sparrows (Zonotrichia leucophrys, Fig. 1), birds that have long been used in this line of research, can assemble a complete song in proper sequence when exposed early in life only to fragments of that song.
Song in birds (and speech in humans) is the culmination of the interplay between innate physiological constraints and environmental cues, including social interactions, song models and auditory feedback. The process is further shaped by developmental constraints such as the limited ‘sensitive periods’ when song models can be memorized. Researchers investigating song learning must therefore assess the degree to which a bird's success or failure to acquire a song results from interactions among these many influences. For example, when presented with tape-recorded songs of their own and other species, juvenile sparrows tend to choose the songs of their own species2. But this innate tendency can be overcome if a bird has a live tutor of another species: if the song structure falls within the capabilities of the sparrows, they can learn to sing the alien songs3.
In general, the morphological structure of song segments (phrases) is a surprisingly weak constraint on song acquisition for white-crowned sparrows. For example, although the trill phrase normally consists only of downward-sweeping frequency modulations, birds can learn songs with upward-sweeping modulations4,5. But what about the order in which phrases are sung? Sparrows tutored on only individual phrases assemble a more complex and species-typical pattern but fail to produce a normal song6.
In their research, Rose et al.1 found a similar limitation. However, they then went on to show that when sparrows were presented with phrase pairs (for example AB, BC, CD, DE or DE, CD, BC, AB) in fixed order, the birds tended to assemble the song correctly (ABCDE). Remarkably, if birds were trained with BA, CB, DC, ED, they tended to learn the song EDCBA. These results show that, for white-crowned sparrows, the acquired template represents sequence information; that the minimal information for identifying a sequence of appropriate length is sufficient to describe song; and that environmentally supplied sequence information can overcome the innate tendency to start songs with whistles (‘A’ in the above examples).
How is this information represented in the brain? How are the early memories of songs laid down, and how do they guide vocal learning? Songbirds possess a set of distinct forebrain cell groups (‘nuclei’) associated with song production, perception and learning — the so-called song system. Early studies of adult birds showed that playback of a bird's own song elicits much stronger responses from song-system neurons than does playback of virtually any other song, a clear result of the learning process. In white-crowned sparrows, some neurons in the nucleus HVC, a site of sensory and motor integration in the song system, exhibited own-song selective responses, but only when a bird was presented with sequences of two phrases drawn in proper order from the bird's song7. The individual phrases failed to stimulate such combination-sensitive neurons, whereas most neurons in the HVC responded to individual phrases5.
Rose and colleagues' results1 imply that an acquired template sufficient for normal song learning can be assembled from sequential pairs of phrases. It is tempting to link these two observations: auditory memories are laid down (in part) as sequences that are represented by combination-sensitive neurons in the HVC. This also requires that such neurons have special ‘status’ as auditory memory neurons (for example, they could be disproportionately common as HVC output neurons), or strongly influence the discharge of the more numerous HVC neurons that respond to single phrases. One caveat is that the observed responses of own-song-selective neurons, recorded in adults, could be the result of establishing early auditory memories, but could also result from practice during vocal development. Evidence from young zebra finches, constrained to sing abnormal songs, shows that at least some song-system neurons achieve selectivity for auditory memories apparently independently of what the bird is singing8.
A related question is whether the acquired template exists as a single group of cells that is modified by early auditory experience, or whether it has a more distributed representation. In the bird forebrain, secondary auditory pathways give rise to at least some of the auditory input to the song system9. Recent evidence, again in adult birds, ties specific changes in auditory responses within these pathways to perceptual learning of songs from different individuals10. If these same pathways are also modified during juvenile acquisition of auditory memories, at the same time that combination-sensitive responses are first specified in the song system, this would be evidence for a distributed representation across multiple cell groups.
Rose and colleagues' studies extend Peter Marler's classic work on white-crowned sparrows that started more than 40 years ago11. They add confidence to the view that the next decade will see dramatic progress in tackling the fundamental physiological questions about the song system. The promise of this system is that research can go beyond addressing which neurons and currents are modified during various phases of learning, to explore how they are arranged into a code for larger units of behaviour and how they produce physiological variation and adaptation. Playing Mother Sparrow to nestlings can be gruelling work, but the new results1 emphasize that behavioural biology is an essential part of this research programme.
Rose, G. J. et al. Nature 432, 753–758 (2004).
Marler, P. A. J. Comp. Physiol. Psychol. 71, 1–25 (1970).
Baptista, L. F. & Petrinovich, L. Anim. Behav. 34, 1359–1371 (1986).
Konishi, M. in Perception and Experience (eds Walk, R. D. & Pick, H. L. J.) 105–118 (Plenum, New York, 1978).
Margoliash, D. J. Neurosci. 6, 1643–1661 (1986).
Soha, J. A. & Marler, P. J. Comp. Psychol. 115, 172–180 (2001).
Margoliash, D. J. Neurosci. 3, 1039–1057 (1983).
Solis, M. M. & Doupe, A. J. J. Neurosci. 19, 4559–4584 (1999).
Vates, G. E., Broome, B. M., Mello, C. V. & Nottebohm, F. J. Comp. Neurol. 366, 613–642 (1996).
Gentner, T. Q. & Margoliash, D. Nature 424, 669–674 (2003).
Marler, P. & Tamura, M. Condor 64, 368–377 (1962).