Miki Takahasi

Stepwise acquisition of vocal combinatorial capacity in songbirds and human infants

Human language, as well as birdsong, relies on the ability to arrange vocal elements in new sequences. However, little is known about the ontogenetic origin of this capacity. Here we track the development of vocal combinatorial capacity in three species of vocal learners, combining an experimental approach in zebra finches (Taeniopygia guttata) with an analysis of natural development of vocal transitions in Bengalese finches (Lonchura striata domestica) and pre-lingual human infants. We find a common, stepwise pattern of acquiring vocal transitions across species. In our first study, juvenile zebra finches were trained to perform one song and then the training target was altered, prompting the birds to swap syllable order, or insert a new syllable into a string. All birds solved these permutation tasks in a series of steps, gradually approximating the target sequence by acquiring new pairwise syllable transitions, sometimes too slowly to accomplish the task fully. Similarly, in the more complex songs of Bengalese finches, branching points and bidirectional transitions in song syntax were acquired in a stepwise fashion, starting from a more restrictive set of vocal transitions. The babbling of pre-lingual human infants showed a similar pattern: instead of a single developmental shift from reduplicated to variegated babbling (that is, from repetitive to diverse sequences), we observed multiple shifts, where each new syllable type slowly acquired a diversity of pairwise transitions, asynchronously over development. Collectively, these results point to a common generative process that is conserved across species, suggesting that the long-noted gap between perceptual versus motor combinatorial capabilities in human infants may arise partly from the challenges in constructing new pairwise vocal transitions.

Ethological data mining: an automata-based approach to extract behavioral units and rules

We propose an efficient automata-based approach to extract behavioral units and rules from continuous sequential data of animal behavior. By introducing novel extensions, we integrate two elemental methods—the N-gram model and Angluin’s machine learning algorithm into an ethological data mining framework. This allows us to obtain the minimized automaton-representation of behavioral rules that accept (or generate) the smallest set of possible behavioral patterns from sequential data of animal behavior. With this method, we demonstrate how the ethological data mining works using real birdsong data; we use the Bengalese finch song and perform experimental evaluations of this method using artificial birdsong data generated by a computer program. These results suggest that our ethological data mining works effectively even for noisy behavioral data by appropriately setting the parameters that we introduce. In addition, we demonstrate a case study using the Bengalese finch song, showing that our method successfully grasps the core structure of the singing behavior such as loops and branches.

Differential androgen receptor expression and DNA methylation state in striatum song nucleus Area X between wild and domesticated songbird strains

In songbirds, a specialized neural system, the song system, is responsible for acquisition and expression of species‐specific vocal patterns. We report evidence for differential gene expression between wild and domesticated strains having different learned vocal phenotypes. A domesticated strain of the wild white‐rumped munia, the Bengalese finch, has a distinct song pattern with a more complicated syntax than the wild strain. We identified differential androgen receptor (AR) expression in basal ganglia nucleus Area X GABAergic neurons between the two strains, and within different domesticated populations. Differences in AR expression were correlated with the mean coefficient of variation of the inter‐syllable duration in the two strains. Differential AR expression in Area X was observed before the initiation of singing, suggesting that inherited and/or early developmental mechanisms may affect expression within and between strains. However, there were no distinct differences in regions upstream of the AR start codon among all the birds in the study. In contrast, an epigenetic modification, DNA methylation state in regions upstream of AR in Area X, was observed to differ between strains and within domesticated populations. These results provide insight into the molecular basis of behavioral evolution through the regulation of hormone‐related genes and demonstrate the potential association between epigenetic modifications and behavioral phenotype regulation.

Developmental learning of complex syntactical song in the Bengalese finch

We developed a neural network model for studying neural mechanisms underlying complex syntactical songs of the Bengalese finch, which result from interactions between sensori-motor nuclei, the nucleus HVC (HVC) and the nucleus interfacialis (NIf). Results of simulations are tested by comparison with the song development of real young birds learning the same songs from their fathers. The model shows that complex syntactical songs can be reproduced from the simple interaction between the deterministic dynamics of a recurrent neural network and random noise. Features of the learning process in the simulations show similar trends to those observed in empirical data on the song development of real birds. These observations suggest that the temporal note sequences of songs take the form of a dynamical process involving recurrent connections in the network of the HVC, as opposed to feedforward activities, the mechanism proposed in the previous model.

A rhythm landscape approach to the developmental dynamics of birdsong

Unlike simple biological rhythms, the rhythm of the oscine bird song is a learned time series of diverse sounds that change dynamically during vocal ontogeny. How to quantify rhythm development is one of the most important challenges in behavioural biology. Here, we propose a simple method, called ‘rhythm landscape’, to visualize and quantify how rhythm structure, which is measured as durational patterns of sounds and silences, emerges and changes over development. Applying this method to the development of Bengalese finch songs, we show that the rhythm structure begins with a broadband rhythm that develops into diverse rhythms largely through branching from precursors. Furthermore, an information-theoretic measure, the Jensen–Shannon divergence, was used to characterize the crystallization process of birdsong rhythm, which started with a high rate of rhythm change and progressed to a stage of slow refinement. This simple method provides a useful description of rhythm development, thereby helping to reveal key temporal constraints on complex biological rhythms.

Trade-offs and correlations among multiple song features in the Bengalese Finch

ABSTRACT A birdsong involves multiple traits that may have evolved under sexual selection pressure. There are two types of song traits: performance-related and elaboration-related traits. These two aspects of songs are partially independent, reflecting neural development and physical condition, respectively, but some song traits might interact with each other because they share the same mechanism for song production. Understanding the evolution of multiple ornaments requires knowledge of correlations among ornaments in the same individual. We explored the potential relationships between the following five song measures. We measured song duration and note rate as performance-related traits; and average note types, linearity index score, and entropy as elaboration-related traits for the analysis. First, we found a significant relationship between linearity and entropy, indicating that syntactical complexity was consistently measured in both different variables. However, note type repertoire was not significantly associated with the two measures of syntactical complexity. Different song nuclei are responsible for each aspect of hierarchically organized song structures. Specifically, a lower-order song nucleus (RA) codes note type, while higher-order song nuclei (HVC and Nif) program transition patterns. Considering that female Bengalese Finches prefer syntactically complex songs, sexual selection, especially female choice, has played a role in shaping brains; however, the sexual selection pressure for each song nucleus may differ in its intensity. Our investigation also revealed that those birds having a larger repertoire of note types tended to sing at lower speeds and require longer song bouts to sing syntactically complex songs. Hence, these results indicate that there are trade-offs and correlations between distinct aspects of performance-related and elaboration-related traits.

Song Motor control organizes acoustic patterns on two levels in Bengalese finches (Lonchura striata var. domestica)

Based on statistical analyses of song sequences, Bengalese finch (Lonchura striata var. domestica) songs do not show unvarying motif repetition as has been found in zebra finches (Taeniopygia guttata). Instead, there are variations of partially stereotyped sequences of song syllables. Although these stereotyped sequences consist of multiple syllables, in most cases these syllables occur together. To examine whether such structures really exist as a vocal production unit, we subjected singing birds to a light flash and determined when the stimulus stopped the songs. When light interruptions were presented within the statistically stereotyped sequences, the subsequent syllables tended to be produced, whereas interruptions presented during the statistically variable sequences tended to cause instantaneous song termination. This suggests that the associations among the song syllables that compose the statistically stereotyped sequences are more order dependent than those for the statistically variable sequences, and the tolerances of syllable pairs to visual interruptions are consistent with the statistical song structures. Additionally, following interruptions, several types of song sequence variations were observed that had not been previously reported. These phenomena might be caused by various effects of the visual stimulus on the hierarchical motor control program.

Domestication changes innate constraints for birdsong learning

Birdsongs are acquired by imitating the sounds produced by conspecifics. Within a species, songs diverge by cultural transmission, but the range of species-specific features is restricted by innate constraints. Bengalese finches (Lonchura striata var. domestica) are a domesticated strain of the wild White-rumped munia (Lonchura striata). The songs of the domesticated strain have more tonal sounds and more variable sequences than those of the wild strain. We compared the features of songs that were produced by normal birds, isolation-reared birds, and cross-fostered birds in both White-rumped munias and Bengalese finches to identify differences in the genetic and environmental factors of their songs. Factor analyses were conducted based on 17 song measurements. We found that isolated songs differed from normal and cross-fostered songs, especially in unstable prosodic features. In addition, there were significant differences in sound property of mean frequency between the two strains regardless of the rearing conditions. Thus, innate constraints that partially determine birdsong phenotypes may be altered through domestication.

Case studies of song and call learning by a hybrid Bengalese–Zebra Finch and Bengalese-fostered Zebra Finches: Assessing innate factors in vocal learning

ABSTRACT Vocal learning in birds is limited by both genetic and environmental factors. Innately determined, species-specific ranges of song parameters restrict song structures; the actual parameter values are determined by environmental inputs during learning. This is also true for some call notes. Here, we investigated the characteristics of songs and distance calls in a hybrid Bengalese–Zebra Finch and Bengalese-fostered Zebra Finches. The Bengalese-fostered Zebra Finches learned a limited part of the song elements, whereas the hybrid learned most of the foster fathers song. However, the hybrid could not learn the acoustic morphology of Bengalese distance calls as well as the fostered finches. The results suggest that the neural circuitry for song element repertoire and call acoustic morphology are independent and that hybridization and cross-fostering procedures are useful tools for investigating genetic and environmental factors in vocal learning.

A Reversible Automata Approach to Modeling Birdsongs

We propose a new automata-based approach to modeling birdsongs on the basis of Angluins induction algorithm, which ensures that k-reversible languages can be learned from positive samples with polynomial time. There are similarities between Angluins algorithm and the vocal learning of songbirds; for example, during a critical period, songbirds also learn songs from positive samples of conspecific birds. Using the proposed method, we demonstrate that the song syntaxes of the Bengalese finch can be represented as reversible automata with lower k-reversibility and that juvenile song syntaxes have two types of development. Our approach provides an effective way to understand the vocal learning of songbirds in terms of computational learning