The study, published in Communications Biology, was designed as a direct, head-to-head comparison between species. Human participants and zebra finches were trained to distinguish sequences of five harmonic tones, artificial stimuli built from the same acoustic ingredients that make up both music and speech. Once both groups had learned the task, the researchers presented modified versions of the sequences in which specific acoustic dimensions had been selectively altered. In some trials the pitch contour, the pattern of how high or low the tones rose and fell across the sequence, was changed. In others the formant contour was shifted, meaning the spectral structure of the tones, the distribution of energy across frequency bands that gives sounds their vowel-like color, was manipulated. In a third condition both dimensions changed together. This factorial design allowed the team to ask, for each species, which of the two dimensions actually drove recognition.

The results revealed a clean and unexpected divergence. Zebra finches treated pitch trajectories and formant trajectories as complementary, interchangeable components of the sound pattern. When one dimension was made less informative, the birds could fall back on the other, and either cue alone was sufficient to support recognition of the sequence. Human listeners behaved very differently. Whenever reliable pitch information was available, they based their judgments overwhelmingly on pitch, and changes to the formant pattern had relatively little influence on how they classified the sequences. In effect, the two species heard the same sounds but filed them under different headings.

Corresponding author Henkjan Honing, professor and director of the Music Cognition Group at the University of Amsterdam, framed the result as a challenge to a popular assumption. Vocal-learning species, a club that includes humans and songbirds but excludes most other mammals, are often assumed to share deep commonalities in how they process sound, on the logic that learning to imitate complex vocalizations imposes similar computational demands. The new findings show that vocal learning does not by itself guarantee a shared listening strategy. How a brain organizes complex sound, Honing notes, appears to depend on two things working together: species-specific auditory predispositions and the attentional habits that experience has taught the listener to apply.

The technical heart of the study lies in the distinction between pitch and spectral structure, two dimensions that are easy to conflate in everyday listening but separable in the laboratory. Pitch corresponds to the fundamental frequency of a harmonic sound, the rate at which the waveform repeats, and it is the primary carrier of melody in music and of intonation in speech. Formants are resonant peaks in the spectrum, shaped by the filtering properties of a vocal tract, and they are the main carriers of vowel identity in human speech. Because the experimental stimuli were synthetic harmonic tones, the researchers could shift formant patterns without changing fundamental frequency, and vice versa, cleanly dissociating the two cues. Most natural listening conflates them; this design pulled them apart and asked each species which thread it chose to follow.

Co-author Zhi-Yuan Ning, who conducted the work at the Institute of Biology Leiden of Leiden University, emphasized that the difference between the species is not a matter of detection. Humans and zebra finches can both perceive changes in pitch and changes in spectral structure. The key difference lies in how those features are used and combined when a listener must recognize a sound sequence. Humans strongly prioritize pitch, treating it as the default dimension along which auditory patterns are organized. Zebra finches weight the two dimensions more evenly, combining pitch with formant information in a way that makes their recognition robust to degradation in either channel. The finding reframes cross-species comparisons of auditory cognition: the interesting question is not which features an animal can hear, but which features its brain elects to trust.

The human data contained a second surprise with implications far beyond the laboratory. Musical experience turned out to modulate performance. When reliable pitch information was removed from the stimuli, listeners with more musical training were markedly better at recognizing the relationships within the sequences by attending to corresponding changes in spectral structure. The response pattern suggests that experienced listeners can map relationships they have learned in the pitch domain onto a different acoustic dimension, recruiting formant-based cues when the familiar cue fails. Musical training, in other words, appears to make auditory processing more flexible. Pitch remains the dominant dimension for human listeners regardless of background, but musical experience seems to widen the fallback options available when pitch is no longer reliable.

For researchers who use songbirds as models of human speech learning, the results carry a cautionary note. Zebra finches have long served as the leading animal model for vocal learning because juvenile birds must hear and memorize a tutor song and then practice it, engaging neural circuits that parallel, in broad outline, those involved in human speech acquisition. But if finches and humans assign different weights to pitch and spectral structure when recognizing the same sequences, then conclusions about human auditory processing drawn from songbird experiments cannot simply be transferred across species. Carel ten Cate, professor at Leiden University and co-author of the study, argues that comparisons of this kind are precisely what allow scientists to distinguish aspects of auditory processing that may be deeply rooted in evolution from those shaped by individual experience and human culture. Both species perceive pitch and spectral information; what differs is the weighting, and that weighting is where biology and culture part ways.

The broader picture that emerges is not a simple biological divide between humans and birds but an interplay between inherited predispositions and learned attention. One plausible reading of the data is that zebra finches are predisposed to integrate several properties of complex sounds, a strategy suited to the demands of song recognition in their natural acoustic world, where individual and species identity are distributed across multiple acoustic features. Humans, by contrast, are immersed from birth in pitch-rich music and speech, environments in which fundamental frequency is the single most informative dimension for melody, intonation and prosody. Extensive experience with these signals may reinforce pitch as the default organizing dimension of human auditory pattern recognition, while leaving enough plasticity, especially in musically trained listeners, to recruit other dimensions when needed.

The implications extend well beyond music. Pitch and formant information are central to speech perception, to animal communication, and to the way any listener parses a complex auditory scene, from a crowded room to a forest at dawn. Understanding how different species organize these dimensions, and how experience reshapes that organization within a single species, contributes to broader theories of auditory cognition, vocal learning and the biological foundations of language and musicality. The zebra finch, it turns out, is not a simplified version of a human listener but a differently tuned one, and the difference itself is the discovery. What humans and songbirds share is the capacity to learn complex vocalizations; what separates them is the grammar their brains impose on the sounds they hear, a grammar written jointly by evolution and by everything each listener has ever heard.

Subject of Research: Comparative auditory sequence recognition in humans and zebra finches

Article Title: Humans and songbirds hear the same sounds, but organize them differently

Article References: Humans and songbirds hear the same sounds, but organize them differently. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: auditory cognition, zebra finch, pitch perception, formants, vocal learning, music cognition, spectral structure, Leiden University, University of Amsterdam, Communications Biology, musical training, songbird model

The research addresses one of the most enduring puzzles in music cognition. Humans do not merely react to beats after they occur; we anticipate them, tapping slightly ahead of or behind the pulse in ways that follow lawful, measurable patterns. Decades of psychophysical experiments have catalogued these patterns in meticulous detail, including the negative mean asynchrony—the tendency for people to tap a few tens of milliseconds before the actual beat—and the way synchronization error grows predictably as tempo increases. What has remained contested is what kind of internal mechanism could give rise to these behaviors.

Traditional computational models have approached the problem in two main ways. Event-based models treat beat tracking as a discrete prediction problem, estimating the time of the next beat from a sequence of previous onsets. Dynamical systems models, by contrast, embed rhythm perception in continuous oscillatory processes, with neural oscillators that entrain to external periodic input. Each family of models captures part of the behavioral data, but each requires engineers to hand-design key structural assumptions about how timing information is represented. The new work asks a different question: if a general-purpose learning system is simply rewarded for staying in sync, will it discover these human-like dynamics on its own?

To find out, the researchers built recurrent neural networks—flexible computational systems whose internal activity depends on both their current input and their own prior states—and trained them using reinforcement learning. Rather than being given labeled examples of correct taps or supervised gradient signals tied to human behavior, the networks received a simple scalar reward for minimizing the error between their produced taps and the underlying beat of rhythmic sequences. This setup mirrors the ecological situation faced by a human learner, who must discover synchronization strategies through feedback rather than explicit instruction about where the beat lies.

After training on a diverse repertoire of rhythmic patterns and tempos, the networks were tested under the same conditions used in classic human tapping experiments. The results were notable on several fronts. The trained networks exhibited a negative mean asynchrony comparable in magnitude to that observed in human participants, and they reproduced the characteristic scaling of synchronization variability with tempo. When the rhythmic stimulus changed speed or switched pattern mid-sequence, the networks showed phase-correction and period-correction behaviors that closely matched human adaptation curves, including the asymmetric way people correct small errors more readily than large ones.

Perhaps most intriguingly, the internal dynamics of the trained networks revealed interpretable structure. Analysis of their recurrent activity showed convergence toward low-dimensional trajectories that effectively encoded beat phase and period, resembling the oscillator-like states posited by dynamical theories of beat perception. The networks had not been told to build internal clocks, yet the pressure of the reward function drove them to develop timing representations functionally equivalent to those long hypothesized in the neuroscience of rhythm. This suggests that human-like synchronization behavior may emerge from general learning principles rather than from rhythm-specific neural hardware alone.

The findings carry implications that extend beyond music cognition. Beat synchronization is increasingly used as a clinical and developmental marker: individual differences in tapping performance have been linked to language development, reading ability, and a range of neurological conditions, including Parkinson’s disease and developmental dyslexia. If a simple reinforcement-trained network can capture the canonical signatures of human performance, researchers gain a tractable model system for probing which aspects of synchronization depend on sensorimotor learning, on neural oscillation, or on the statistics of the rhythms themselves. Perturbing the artificial networks—lesioning units, altering reward structure, or changing input statistics—offers a level of experimental control impossible with human participants.

The work also speaks to a broader debate in artificial intelligence and neuroscience about the explanatory power of trained neural networks as scientific models. Critics argue that deep learning systems can fit behavior without illuminating mechanism. This study pushes back on that concern by showing that a behaviorally constrained network not only matched human outputs but also reproduced interpretable internal dynamics aligned with theoretical predictions. In this sense, the network functions as a computational hypothesis: it demonstrates that the reward structure of a synchronization task is sufficient, in principle, to generate the observed suite of human tapping phenomena, narrowing the space of explanations that cognitive science needs to consider.

Limitations remain, as the authors acknowledge. Real-world beat perception unfolds in acoustically rich environments, with polyphonic music, expressive timing, and social coordination among multiple players, whereas the trained networks operated on more controlled rhythmic stimuli. Human synchronization also engages motor physiology—the biomechanics of the limb, spinal circuitry, and cortical motor areas—that a simulated tapping output does not capture. Extending the approach to multimodal, embodied settings will be an important next step, as will testing whether networks trained on different reward schedules produce different asynchrony profiles, which could help explain the substantial individual variability seen in human tapping data.

Even with those caveats, the study marks a compelling convergence between machine learning and cognitive science. It suggests that the human sense of the beat—so effortless that we rarely notice it—may be the natural solution to a general predictive-control problem, one that an artificial learner rediscovers when faced with the same objective. As reinforcement-trained models continue to reproduce facets of perception, action, and now rhythmic coordination, the boundary between engineered and biological intelligence grows both more interesting and more informative to study.

Subject of Research: Reinforcement-trained recurrent neural networks as computational models of human beat synchronization and sensorimotor timing

Article Title: Reinforcement-trained recurrent networks reproduce human beat-synchronization dynamics

Article References: Ommi, Y., Yousefabadi, M., & Cannon, J. (2026). Reinforcement-trained recurrent networks reproduce human beat-synchronization dynamics. Communications Psychology. https://doi.org/10.1038/s44271-026-00529-1

Image Credits: AI Generated

DOI: 10.1038/s44271-026-00529-1

Keywords: beat synchronization, recurrent neural networks, reinforcement learning, music cognition, sensorimotor timing, rhythm perception, neural oscillators, tapping behavior, computational neuroscience, predictive timing, Reinforcement-trained, recurrent