Z Zebra Song
Lahoma Jenkins
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
We thank J. Wallman and H. Williams for critical reading of the manuscript and consultation and F. Nottebohm, C. Harding and N. Leader for recordings of WT songs. The study was supported by US National Institutes of Health (NIH) grants to O.T. and P.P.M., by a NIH Research Centers in Minority Institutions grant to City College, City University of New York, and by the Crick-Clay Professorship to P.P.M.
Author Contributions The idea for the study originated with P.P.M., with important modifications by O.T. and O.F. The experiments were carried out by O.F. and O.T. The model was developed by P.P.M. with help from H.W. All authors participated in the data analysis, with major efforts by H.W. and O.F.
Vocal culture is the cornerstone of spoken language, but is not unique to humans1,2,3,4,5. Like humans, songbirds acquire their vocal repertoire via imitation (i.e., vocal learning)6,7,8,9, a process that can give rise to local dialects that persist over hundreds of generations10,11. However, the repertoire of vocal learning birds also has a strong genetic component11,12,13. Across populations, innate biases in song perception, production, and learning sustain species-specific song repertoires13,14,15,16. Canaries, for example, will faithfully imitate songs of abnormal combinatorial structure, but later, as they reach maturity, alter their songs to match a species typical song syntax to which they have not been exposed17. Similarly, zebra finch males (females do not sing) that are trained with random combinatorial transitions of syllable types will generate combinations that are biased toward the species typical18,19. Innate biases may unfold at the scale of generations, too; the descendants of isolated zebra finch tutors, who produce aberrant songs, produce increasingly species typical songs2,19.
Theoretically, vocal imitation should drive song repertoire convergence within groups and divergence across groups20,21,22. Meanwhile, innate biases in imitation might constrain drift. In reality, however, zebra finch songs remain highly diverse within groups and vary only mildly across them22. We do not know whether this diversity serves any function in domesticated zebra finches, but high similarity between songs could potentially generate impoverished communication systems that convey little information about individual identity23,24. In wild songbirds, across species, and even subspecies, the magnitude of individual song variability differs strongly, often for no apparent reason. For example, the songs of the wild Australian zebra finch (Taeniopygia guttata castanotis) are much more variable among individuals than those of the closely related wild Timor zebra finch (Taeniopygia guttata guttata)25. This variability persists despite the fact that they live in similar climates and have similar social organization.
Here we test how a rich polymorphic repertoire of song syllables is sustained during cultural transmission26 in the Australian zebra finch. We quantify song polymorphism using novel measures of vocal states and acoustic diversity, for studying the statistics of song imitation in a large colony. We find that the polymorphic repertoire is sustained by pupils spontaneously increasing song diversity when tutors have low-diversity songs, and imitating with greater fidelity when tutors have high-diversity songs, a process we call balanced imitation.
a 24 song tutoring lineages. All tutors had pupils in more than a single clutch. Each node represents one individual animal. Node shape represents pupils from the same clutch. Tutor nodes are presented on the bottom and pupil nodes on the top. Similarity scores are presented as quartiles (green for best imitations and red for poorest). Lineages are sorted according to the mean similarity between tutor and pupils from highest (top) to lowest (bottom). b, c Examples of song imitations from tutor AQ12 with a low similarity family (b) and from tutor DG1 with a high similarity family (c). Imitation outcomes are presented as percent acoustic similarity estimates on each sonogram. Red bars outline the repeated song motifs of the tutors. Source data for this figure is in Supplementary Data File 1.
Our measures up to now summarize the distribution of vocal states within a song. We next looked at each vocal state separately and measured how frequencies (abundances) of vocal states are imitated. In prior studies, we noted that vocal imitation in zebra finches is inversely related to model abundance. That is, too much exposure to a tutored song could inhibit learning31. Here we test if this is the case also for abundances of vocal states within a song.
We superimposed these empirically determined pitch intervals (for top and bottom influence) on ranges of mean song pitches obtained in a database of four zebra finch colonies including the current one, and shaded the intervals values green (presumably top influence) and red (presumably low influence; Fig. 7f). We then did the same for frequency modulation (Fig. 7g), and Weiner entropy (Fig. 7h). Across the colonies, the distribution of mean song features was to a large extent confined within the range of high influence in our colony. Therefore, the range of mean feature values of highest imitation influences in our colony, but not of lowest influences, seems consistent across zebra finch colonies. This range, in turn, can be explained by balanced imitation as high influences are associated with high tutor song diversity. In sum, this outcome is consistent with the notion that over generations, songs of high feature diversity are more influential, and therefore shape the overall distribution of mean song features in a similar manner across colonies.
It would be interesting to test if balanced imitation parameters are different across species. Variation in the intensity of the trend to sustain high song diversity and that of the trend to imitate songs accurately could lead to equilibriums that differ according to species and possibly even the ecological conditions in which a species lives. Perhaps species with songs that are similar across individuals engage in weak balanced imitation and vice versa. For example, to explain why the songs of the Timor zebra finch are much more similar across individuals compared to the Australian zebra finch25, we speculate that perhaps a weaker balanced imitation gain in the Timor zebra finches (compared to Australian zebra finches) could potentially increase the odds of extinction of rare song elements, driving the stronger convergence observed in songs across individuals.
Regardless of possible prevalence across species, accounting for balanced imitation in zebra finches might be necessary in order to properly interpret vocal learning outcomes. This is particularly important because mechanisms of vocal learning are studied extensively in Estrildid finches, among which song learning outcomes vary considerably across individuals. In part, this variability is associated with factors like genetics and with tutoring mismatches12,36. Our results indicate that, in addition, deviations from tutor song through reorganization and transformation of copied vocal sounds may be driven by an inclination to optimize song diversity. This can be regarded as a discrete form of error correction during song learning. That is, balanced imitation involves correcting errors from states of minimal (and perhaps also maximal34) diversity. In the framework of error correction37, the developmental question is when and how the vocal learning bird balances between error correction exclusively in reference to tutor sound to error correction in reference to a state of its own sound diversity. A better understanding of this balance and possible transition could reveal the mechanism through which a species-specific level of cultural song diversity is determined23.
Another observation that requires further study is the recombination of syllable units. In the cases we observed, pupils combined tutored syllables into new and more complex units, but splitting appears to be rare. Splitting could be an artifact due to limitations of our methods in detecting such recombinations; it could also suggest a tendency to compress the tutored song. Such a compression might be useful when several potential tutors are available. Compressing the imitation from one song could leave more room to imitate song elements from other tutors. Further, some improvised syllable types tend to be acoustically simple and are transformed across generations into complex types2. This line of thinking suggests that perhaps we should consider not only the overall acoustic diversity of a tutored song, but also the diversity per unit time. Here too, variation across species can be potentially explained: as opposed to zebra finches, in Bengalese finches, syllable level analysis shows correlation in song and transition diversity across tutors and pupils38. We do not know if syllable recombination is common in Bengalese finches, but they usually produce less complex syllable types compared to zebra finches, which could suggest that Bengalese finches are less inclined to compress their songs.
Previous studies25,39 suggest that high song diversity in a colony of zebra finches could be adaptive. In the zebra finch female, brain dopamine response to songs is tuned to the song of her mate39. To the extent that balanced imitation can also sustain the acoustic diversity of songs within a colony, it might also enable the females to respond selectively to their mates. Balanced imitation is also of interest in a broader context of vocal and non-vocal cultures in humans. In general, cultures may vary in their stability and in their richness (polymorphism), and balanced imitation could potentially explain how different morphs of cultures come about. At the population level, balanced imitation can be thought of as an example of a balancing (negative frequency-dependent) selection of morphs, which can promote polymorphism by preventing the extinction of rare morphs. At the individual level, it can be thought of as a mechanism that promotes diversity in the skills that are acquired. It would be particularly interesting to test how imitation biases might interact with the structure (topology) of communication networks, in determining how cultural behavior spreads and is filtered over space and time40. Finally, other possible mechanisms could potentially explain balanced imitation including perceptual biases41 and habituation42.
c80f0f1006