Humans, but not zebra finches, produce spontaneously isochronous song through the lifespan Authors: Mila Bertolo, Michael W. Weiss, Isabelle Peretz, Jon T. Sakata Published: 2026 ABSTRACT SummaryIsochrony—the regular beat of a metronome—is cross-culturally ubiquitous in music. Is this ubiquity due to a widespread biological inclination for isochrony in acoustic communication? If so, it should be present in lesser-studied vocal music, in the absence of musical training and early in development, and in comparable non-human species’ vocalizations. We quantified isochrony in improvised songs from non-musician adults (n = 15, 24–82 years old) and children (n = 38, 3–10 years old), and in songs from juvenile and adult zebra finches (n = 77, ∼50 and >120 days post-hatch). Isochrony was expressed in Western non-musician and children’s improvised songs to a comparable degree. In contrast, both juvenile and adult zebra finches produced songs with less isochrony than chance, with no evidence for the learning of isochrony. These data show that spontaneous isochrony in learned vocalizations appears differently in two species. We propose that species variation in chorusing behaviors may explain why. Introduction Rhythm is one of the basic building blocks of human musicality.1 In its simplest form, the use of a regular pulse—like the clicking of a metronome, referred to as isochrony—is extremely widespread among music traditions.2,3 Musicians from Mali, Bulgaria, and Germany all finger-tap to reliably mimic small integer ratio rhythms, where durations of adjacent notes have small integer ratio relationships, such as 1:1 and 1:2; they also show a tendency to simplify complex rhythms (e.g., to 1:1, isochrony).4 This ubiquity suggests the behavior has some shared, biological basis. Indeed, the ability to clap in time to a perceived isochronous pulse has an identifiable polygenetic basis in humans.5 The presence of isochrony in the vocalizations of distantly related non-human animal species—non-human primates,6,7,8,9 birds,10,11,12 frogs,13 and pinnipeds14,15,16—further indicate some biological basis to this behavior. To date, however, we have little data concerning whether isochrony is expressed in an ancient and universal musical form—singing—and whether there are comparable expressions of isochrony in non-human species capable of combinatorial, learned vocalizations (as opposed to simple, repeated, innate ones). Here, we present a developmental and comparative analysis of isochrony in human improvised songs, and spontaneous song from a non-human animal capable of learned, complex songs: the zebra finch. Isochrony production is widespread and consistent in human percussive musical behaviors. Musicianship is not a prerequisite: non-musicians’ accuracy in tapping synchronously to isochronous stimuli is on par with that of amateur musicians.17 When musicians and non-musicians from 15 countries were tasked with reproducing an auditory rhythm by tapping along to it, their “errors” systematically favored isochrony even when it was not originally present in the stimulus.3 Percussive isochrony emerges early in human development; by 2–4 years old, children’s manual tapping spontaneously show regular rates,18 and they show full-body rhythmic movements in response to music, though inconsistently synchronized to it.19 By 6–7 years old, the ability to synchronize to an isochronous stimulus is almost at adult-level accuracy,20 and this accuracy is well-preserved through aging.21 If isochrony is as fundamental to human musicality as the above findings suggest, we should also see this rhythm in untrained vocal song. Singing is a universal22 and spontaneous behavior in humans, and its perception is associated with specific neural networks.23 Young children create new tunes and adapt familiar ones,24,25 and this practice of real-time improvisation is found across many of the world’s musical traditions.26 Improvisational singing, furthermore, is a methodologically useful tool in that it can be intuitively performed by people with little to no formal musical training, through the lifespan. Western non-musicians’ improvised songs remarkably show pitch use that is consistent with the conventions of Western music.27 Previous findings of isochrony in corpora that include both instrumental and vocal music2 suggest that isochrony is likely also present in song, but such corpora include music that is the product of explicit musical training. Instead, the study of non-musicians’ improvised song—in the absence of explicit training or mimicry—can better characterize how spontaneously isochrony may arise in human musical production. We expect that non-musicians’ improvised songs should show isochrony in line with their percussive musical behaviors. Similarly, this trait should emerge around the age of 6 or 7 in children’s improvised songs. Sung improvisation, which by design imposes very little constraints on participants, also allows for more direct cross-species comparisons. Very few non-human species “improvise”; many species’ vocalizations consist of innate calls that are repeated at consistent and stereotyped intervals13,28 creating, by definition, isochronous sequences. By contrast, vocalizations that combine and sequence many different sound elements do not necessarily result in an isochronous rhythm; we know much less about whether isochrony is also the prevailing rhythmic organization in the combinatorial vocalizations of non-human animals. Songbirds like the zebra finch are ideal for investigating isochrony in learned vocalizations that are combinatorial in nature. Zebra finches learn their song by memorizing and imitating a tutor’s song, and mechanisms underlying the perception, acquisition, and performance of these songs have been extensively studied.29 Juvenile zebra finches’ songs are more temporally variable than mature song30,31; through development, they learn the sequencing and timing of song elements (“syllables”),29,32 affording them a level of temporal control that in principle could be used to give their complex vocalizations rhythmic structure similar to human song. Timing is ecologically relevant for these songbirds; they coordinate their song with body and head movements,33,34 and show inter-individual coordinated timing,35 but notably they do not sing synchronously. Zebra finches are also able to perceptually discriminate between isochronous and non-isochronous stimuli with some degree of tempo flexibility.11 Here, we provide a developmental and comparative analysis of isochrony in humans and songbirds, focusing on spontaneous/improvised songs in each species. We tested for the presence of isochrony in the improvised songs of children and adult non-musicians, and in the spontaneous songs of juvenile and adult zebra finches. While two previous studies have reported isochrony in adult zebra finch song,10,36 no study to date has examined developmental changes in isochrony, analyzed the contribution of learning to isochrony, and directly compared isochrony in zebra finch song to isochrony in human vocal music. This comparative study of isochrony will provide a more detailed account of the degree of similarity between human and songbird vocal rhythms, which will help evaluate candidate theories for the origins of rhythm, including the vocal learning hypothesis for beat perception and synchronization,37 the gradual audiomotor evolution hypothesis,38 and the origins of rhythm production in group dancing39 or other group behaviors.40,41 Results Songs were semi-automatically segmented to timestamp the onsets of notes (in human song) and syllables (in zebra finch song; Figure 1), and dyadic interval ratios were computed based on inter-onset intervals (see STAR Methods).Figure 1Example sound waveformsWaveforms of (A) an adult non-musician's song, (B) a child's improvised song, (C) an adult zebra finch’s song, and (D) song from that same zebra finch as a juvenile. Visually, the intervals between the onsets (green dashed lines) of adjacent notes/syllables are roughly equivalent in human songs—indicating isochronous organization, and more variable in the zebra finch song. We used the dyadic interval ratio to quantify isochrony in the song recordings of human children and adults, and zebra finch juveniles and adults. This is the ratio for any two adjacent inter-onset intervals IOInIOIn+IOIn+1, where a highly regular signal such as a metronome with identical IOIs would only result in ratios of 0.5. Each panel's scale bar indicates 500 ms. Example sound waveforms Waveforms of (A) an adult non-musician's song, (B) a child's improvised song, (C) an adult zebra finch’s song, and (D) song from that same zebra finch as a juvenile. Visually, the intervals between the onsets (green dashed lines) of adjacent notes/syllables are roughly equivalent in human songs—indicating isochronous organization, and more variable in the zebra finch song. We used the dyadic interval ratio to quantify isochrony in the song recordings of human children and adults, and zebra finch juveniles and adults. This is the ratio for any two adjacent inter-onset intervals IOInIOIn+IOIn+1, where a highly regular signal such as a metronome with identical IOIs would only result in ratios of 0.5. Each panel's scale bar indicates 500 ms. Adult non-musicians’ improvised songs display isochrony Adult non-musicians’ improvised songs showed significantly more dyadic interval ratios in the isochrony window than the off window (Figure 2A; n = 15; V = 120, p < 0.001). Additionally, the degree of isochrony in these songs was greater than one would observe by chance (Figure 2B; V = 103, p = 0.012). There was no linear relationship between adults’ ages and proportion of ratios in the isochrony window (β = −0.0001, SE[β] = 0.0058, z = −0.018, p = 0.986).Figure 2Non-musician adult and children’s improvised songs show isochronyDistributions of dyadic interval ratios of (A) adult non-musicians’ songs show more ratios in the isochrony window (“in 0.5”, window highlighted in in orange) than outside of it (“off 0.5”, window highlighted in yellow). Each colored line is an individual participant’s distribution; the thicker black line is the population average. The inset shows each of the 15 adults’ difference in how many ratios fell in 0.5 vs. off 0.5; this is quantified as a normalized count, which is the number of ratios that fell in each window, divided by the width of the windows. Plotted are individual data points and boxplots depicting the median and quartile of the distributions.(B) Adults’ songs show more ratios inside this window than chance. Each blue line plots the distribution of dyadic intervals in one of the 50 scrambled datasets; each point in the boxplot is a participant’s normalized count of ratios in the isochrony window, in the empirical data or averaged in the scrambled data. The inset plots the normalized empirical (observed) and scrambled (chance) counts for each individual. Plotted are individual data points and boxplots.(C and D) Children’s (n = 38) improvised songs show the same pattern as adults. One child’s distribution of ratios is not plotted for clarity (the peak height would visually distort the rest of the graph; see Figure S1). Insets are similar to those plotted in (A) and (B).(E) There is no significant linear relationship between the children’s age and amount of isochrony in their song. Plotted is the regression line with the shaded region indicating the standard error of the estimate.(F) There is no significant difference between children and adults in the amount of isochrony in their song. Plotted are individual data points as well as boxplots depicting the median and quartiles of the distributions. ∗p < 0.05, ∗∗p < 0.01. Non-musician adult and children’s improvised songs show isochrony Distributions of dyadic interval ratios of (A) adult non-musicians’ songs show more ratios in the isochrony window (“in 0.5”, window highlighted in in orange) than outside of it (“off 0.5”, window highlighted in yellow). Each colored line is an individual participant’s distribution; the thicker black line is the population average. The inset shows each of the 15 adults’ difference in how many ratios fell in 0.5 vs. off 0.5; this is quantified as a normalized count, which is the number of ratios that fell in each window, divided by the width of the windows. Plotted are individual data points and boxplots depicting the median and quartile of the distributions. (B) Adults’ songs show more ratios inside this window than chance. Each blue line plots the distribution of dyadic intervals in one of the 50 scrambled datasets; each point in the boxplot is a participant’s normalized count of ratios in the isochrony window, in the empirical data or averaged in the scrambled data. The inset plots the normalized empirical (observed) and scrambled (chance) counts for each individual. Plotted are individual data points and boxplots. (C and D) Children’s (n = 38) improvised songs show the same pattern as adults. One child’s distribution of ratios is not plotted for clarity (the peak height would visually distort the rest of the graph; see Figure S1). Insets are similar to those plotted in (A) and (B). (E) There is no significant linear relationship between the children’s age and amount of isochrony in their song. Plotted is the regression line with the shaded region indicating the standard error of the estimate. (F) There is no significant difference between children and adults in the amount of isochrony in their song. Plotted are individual data points as well as boxplots depicting the median and quartiles of the distributions. ∗p < 0.05, ∗∗p < 0.01. Children’s improvised songs also display isochrony Similar to adults, children aged 3–10 years old spontaneously produced isochrony in their improvised songs (Figure 2C; n = 38; V = 736, p < 0.001), at a higher-than-chance rate (Figure 2D; V = 686, p < 0.001). Additionally, there was no evidence for an influence of age on the degree of isochrony expressed: there was no linear relationship between children’s age and proportion of ratios that fell in the isochrony window (Figure 2E; β = −0.0512, SE[β] = 0.0711, z = −0.721, p = 0.471), and no significant difference between children and adults in proportion of ratios in the isochrony window (Figure 2F; β = −0.045, SE[β] = 0.144, z = −0.312, p = 0.755). Zebra finch song is less isochronous than chance through the lifespan Songs from our large colony sample of semi-naturalistically reared adult zebra finches (n = 61) tended to show more dyadic interval ratios in the isochrony window than in the “off” window, but this comparison was not statistically significant (Figure 3A; V = 1158, p = 0.074). Further analysis revealed that the magnitude of this peak in the isochrony window was significantly lower than would be expected by chance (Figure 3B; V = 191, p < 0.001).Figure 3Adult zebra finch song is less isochronous than chance(A) Each colored line represents the distribution of dyadic interval ratios of an individual adult zebra finch. The thicker black line indicates the population average distribution of dyadic interval ratios. There are not significantly more ratios within the isochrony window (“in 0.5”) than outside it (“off 0.5”). Plotted in the inset are the normalized counts for in 0.5 vs. off 0.5 for each bird (see also Figure 2). Boxplot summarize the median and quartiles of the distributions.(B) The degree of isochrony in zebra finch song was significantly lower than expected by chance. The thicker black line is the same population average as in (A), and each blue line is the distribution of dyadic interval ratios by chance, as reconstructed from each of the 50 scrambles of these same birds’ data. Plotted in the inset are normalized empirical (observed) and scrambled (chance) counts for each bird. Normalized counts are computed by taking the number of ratios observed in a window (“in 0.5” or “off 0.5”) and dividing that by the width of the window. Chance degrees of isochrony are computed from iterations of scrambling the order of each bird's inter-onset intervals and recomputing DIRs. Boxplot depicts median and quartiles. ∗p < 0.05; ∗∗p < 0.01. Adult zebra finch song is less isochronous than chance (A) Each colored line represents the distribution of dyadic interval ratios of an individual adult zebra finch. The thicker black line indicates the population average distribution of dyadic interval ratios. There are not significantly more ratios within the isochrony window (“in 0.5”) than outside it (“off 0.5”). Plotted in the inset are the normalized counts for in 0.5 vs. off 0.5 for each bird (see also Figure 2). Boxplot summarize the median and quartiles of the distributions. (B) The degree of isochrony in zebra finch song was significantly lower than expected by chance. The thicker black line is the same population average as in (A), and each blue line is the distribution of dyadic interval ratios by chance, as reconstructed from each of the 50 scrambles of these same birds’ data. Plotted in the inset are normalized empirical (observed) and scrambled (chance) counts for each bird. Normalized counts are computed by taking the number of ratios observed in a window (“in 0.5” or “off 0.5”) and dividing that by the width of the window. Chance degrees of isochrony are computed from iterations of scrambling the order of each bird's inter-onset intervals and recomputing DIRs. Boxplot depicts median and quartiles. ∗p < 0.05; ∗∗p < 0.01. In the longitudinally recorded birds (n = 16), zebra finches showed significantly more ratios in the isochrony window than in the off window when they were adults but not when they were juveniles (Figures 4A and 4B; juveniles, t[15] = 1.781, p = 0.095; adults, V = 117, p = 0.009). While in isolation this could be indicative of a developmental trend for the emergence of isochronous rhythm, further analysis shows a pattern similar to the colony bird population: this incidence of ratios in the isochrony window was significantly lower than would be observed by chance when birds were both juveniles (t[15] = −2.289, p = 0.037) and adults (V = 20, p = 0.011; Figure 4). There was no overall effect of age on the incidence of isochrony production (β = −0.08, SE[β] = 0.215, z = −0.392, p = 0.695).Figure 4Zebra finch song is less isochronous than chance through the lifespan(A and B) Distributions of dyadic interval ratios from recordings of the same birds when they were (A) juveniles and (B) adults. Zebra finches showed a significantly higher number of ratios in the isochrony window (“in 0.5”) than outside it (“off 0.5”) when they were adults but not when they were juveniles (see Figure 3 for description of insets).(C and D) Both the juvenile and adult songs of these birds showed less isochrony than would be expected by chance (see Figure 3 for description of insets).∗p < 0.05. Zebra finch song is less isochronous than chance through the lifespan (A and B) Distributions of dyadic interval ratios from recordings of the same birds when they were (A) juveniles and (B) adults. Zebra finches showed a significantly higher number of ratios in the isochrony window (“in 0.5”) than outside it (“off 0.5”) when they were adults but not when they were juveniles (see Figure 3 for description of insets). (C and D) Both the juvenile and adult songs of these birds showed less isochrony than would be expected by chance (see Figure 3 for description of insets). ∗p < 0.05. Zebra finches do not appear to learn isochrony naturally We found no evidence for the learning of isochrony: the amount of isochrony (even if below chance) in the songs of tutor birds was not significantly predictive of the amount of isochrony in their pupils’ songs (Figure 5; β = 1.27, SE[β] = 0.9, z = 1.48 p = 0.138).Figure 5No significant relationship between isochrony in a tutor and its pupil’s song in the colony sample of zebra finchesThe amount of isochrony in a tutor’s song is not significantly related to the amount of isochrony in its pupil’s song, suggesting that zebra finches do not learn to produce this rhythm. Plotted are data for individual tutor-pupil pairs, with the line and shaded region indicating, respectively, the regression line and the standard error of that estimate. No significant relationship between isochrony in a tutor and its pupil’s song in the colony sample of zebra finches The amount of isochrony in a tutor’s song is not significantly related to the amount of isochrony in its pupil’s song, suggesting that zebra finches do not learn to produce this rhythm. Plotted are data for individual tutor-pupil pairs, with the line and shaded region indicating, respectively, the regression line and the standard error of that estimate. Discussion We studied rhythm in naturalistic, spontaneous vocalizations from humans and zebra finches. While both species are precocious singers and dependent on social and sensory experiences to learn their songs, only humans’ songs were isochronous, and this species difference persisted through development (in the age ranges considered here). These findings for improvised human song expand on the ecological validity of corpus2 and tapping3 studies that indicate how common isochrony is in human musicality. Our findings also replicated a report that zebra finch song shows a high density of dyadic interval ratios around 0.5.10 However, subsequent analyses indicated that the magnitude of this peak was significantly lower than expected by chance, in both juvenile and adult birds’ songs; consequently, zebra finch song are not isochronous in a way that is analogous to human song. What underlies this species difference is unknown. It is possible that humans, but not songbirds, produced isochrony due to their greater experience with this rhythm, but two of our findings weigh against this possibility. First, if more isochrony exposure in humans led to more isochrony production, we would have observed a developmental increase in isochrony production: this was not the case in our data, with isochrony production being rather constant over the 3–10 years age range, and between children and adults. If isochrony production is experience dependent, this effect might be acting only very early in development. While these participants had minimal musical training, future cross-cultural work may determine how passive musical exposure may influence isochrony production. Second, if isochrony production were a function of exposure, we would have observed that prevalence of isochrony in a zebra finch pupil’s song would mirror that of its tutor’s. This also was not borne out in the data. The lack of correlation in the degree of isochrony between tutor and pupil songs suggests a minimal contribution of learning and is surprising given previous studies of temporal learning. This may be due to birds prioritizing the reproduction of low-level temporal features (i.e., syllable structure and gap durations) over global rhythmic structures (as previously suggested in some studies42,43,44). We observed instances where tutors and pupils showed generally similar sequencing of sounds, but where pupils merged or split syllables from their tutor’s song,45 leading to altered inter-onset intervals and, consequently, dyadic interval ratios. Any isochrony observed in zebra finch song might therefore be incidental to their reproduction of low-level temporal features; this is in contrast to human musical rhythms where the production of an isochronous rhythm is persistent through variation in low-level features like average inter-onset interval (see Figure S2). Future studies among denizens of different musical cultures would help characterize whether the present results are species-typical of humans. The method used here—of recording children’s songs online, supervised by their parents—is a scalable alternative to in-lab methods. This scalability, in addition to this method’s requirement for minimal participant instruction, lends itself to facilitating cross-cultural study. If humans have no predisposed rhythm inclinations and isochrony production is purely experience-dependent, we might see, for instance, Malian children’s improvised songs would show a prevalence of isochronous and non-isochronous rhythms in accordance with their prevalence in jembe music.46 Replications of this work would also reveal whether the lack of an age effect persists in other samples. We found no evidence for an increase in spontaneous isochrony production with age, which was surprising given that body movements to music19 and manual tapping20 become more accurate through the age range of ∼2–7 years old. If replicated, the vocal production of isochrony may be more precocious than previously expected. In addition to the note/syllable level isochrony considered here, isochrony may occur at different time scales across species. For example, the Australian pied butcherbird’s song shows isochrony not at the syllable but at the phrase level, with inter-phrase-onset intervals being of consistent duration.47 We used a prevalent, semi-automated syllable-based segmentation for zebra finch song, where adjacent elements are delineated with brief silences. This is the same method used in prior investigations of isochrony in birdsong10,11 as well as countless other analyses of zebra finch song. Alternatively, one could consider that syllables in bird song can consist of several “gestures,” where gesture onsets have been reported to align with isochronous events more than would be expected by chance36 (though this study also reports isochrony at the syllable level) and to contain some signature of learning.48 Future work can make explicit at what level (gestures, notes, phrases) we expect to observe isochronous organization, in which species, and why. If isochrony’s function, as a perfectly predictable rhythm, is to serve inter-individual coordination or synchronization,49 naturalistic interactions between individuals should show coordination at the levels of segmentation considered. This interpretation is aligned with findings that isochrony is a defining feature of vocalizations that we chorus in, like song, and less so of turn-taking vocalizations like speech41,50 or zebra finch song, where these birds do not chorus, and in fact deliberately avoid overlapping with an interlocutor.42 This species difference has broader implications about the biological basis of musicality. Much ongoing work seeks to determine which species produce rhythmic vocalizations, and why. One popularly evoked condition is vocal learning.6,14,51,52 The vocal learning hypothesis37 suggests that vocal learning neural circuitry is necessary for beat perception and synchronization, and many have tested whether vocal learner species also produce rhythmic vocalizations. For example, a recent study documents that the warble songs of budgerigar, a species of parrots, contain significant isochrony (compared to chance).12 Our data suggest that vocal learning might not be strongly linked to the production of isochronous rhythms in learned vocalizations across species. Future work testing what other features selectively covary with rhythmic vocalizations may elucidate whether music’s rhythms may be attributable to higher vocal learning capacities, or other, yet untheorized species traits. Limitations of the study We used automated procedures that are established and routinely used in the fields to segment human and finch songs. However, the algorithms differ slightly across the species, with pitch information taken into consideration for human but not finch song. We do not believe that this impacts the conclusions of our study because segmenting human songs based purely on amplitude led to similar conclusions about isochrony (Figure S3). Attempts to segment zebra finch song using the software for segmenting human songs were unsuccessful (which is likely due to the substantive differences in acoustic properties of vocal units across species). Other studies have manually segmented zebra finch song into notes (i.e., gestures) and documented evidence of isochrony and learning of temporal patterns,36,48 and this segmentation difference, as well as different analysis approaches for the quantification of rhythm, could lead to different conclusions to ours. Dyadic interval ratios are one of a number of different measures of rhythm, and it is possible that we would find stronger evidence of rhythm learning in finches using other metrics. A greater range of children and a global population of individuals with less musical training could improve our understanding of the biological foundations (e.g., innateness) of rhythm production in humans. Resource availability Lead contact Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Jon T. Sakata (jon.sakata@mcgill.ca). Materials availability This study did not generate new unique reagents. Data and code availability •Data are available as supplementary materials, and publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX.•Code is also publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX.•Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request. •Data are available as supplementary materials, and publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX. Data are available as supplementary materials, and publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX. •Code is also publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX. Code is also publicly archived at OSF: https://doi.org/10.17605/OSF.IO/FN2MX. •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.