Much like human babies, newborn monkeys cry and coo to get their caregiver’s attention. They all produce these sounds in the same way. They push air from the lungs to vibrate the vocal cords, and adjust the movement of their jaws, lips, tongue and other muscles to create different kinds of sounds.

Ultrasounds show that human fetuses begin making crying-like mouth movements during the last trimester of pregnancy. Yet the prenatal development of this crucial skill remains unclear, as most studies of early primate vocalization take place after birth.

To explore this question, Narayanan et al. focused on a small species of monkeys known as marmosets. Regular ultrasounds were performed on four pregnant marmosets, starting on the first day the fetuses’ faces became visible and ending the day before delivery. The developing marmosets acquired the ability to independently move their mouth from their head over time, a skill crucial for feeding and vocalizing. By the end of pregnancy, a subset of fetal mouth movements were nearly identical to those produced when baby marmosets call for their caregivers after birth.

Human ultrasound studies are needed to confirm whether vocal development follows a similar trajectory in our species.This is likely given the developmental similarities between both species. If so, work in marmosets could be helpful to understand how conditions such as cerebral palsy interfere with this process, and to potentially develop early interventions.

Neonatal primates produce vocalizations (e.g. cries and coos) as a rhythmical series of sounds (Hopkins and von Wulfften Palthe, 1987; Wolff, 1969). These vocalizations are often the substrate for the rhythmic vocalizations produced by adult humans (Chandrasekaran et al., 2009) and other primates (Risueno-Segovia and Hage, 2020; Lameira et al., 2015; see Ghazanfar and Takahashi, 2014 and Bergman et al., 2019 for reviews). Their production requires coordination among all the elements of the vocal apparatus, i.e., between laryngeal tension, respiration, and orofacial movements. We know that postnatal vocal production is a complex motor act that changes over time and is influenced by caregiver feedback (Goldstein and Schwade, 2008; Warlaumont et al., 2014) and the ambient environment (Mampe et al., 2009). But how do those very first neonatal vocalizations develop?

During human pregnancies, movements consistent with crying have been observed during ultrasound assessments of fetuses in their last trimester (Gingras et al., 2005) but their prenatal developmental trajectory and differentiation from other movements have not been measured. We reasoned that a nonhuman primate model of this behavior would be a way to gain insights. Marmoset monkeys are a good candidate model for human infant vocal behavior. The early postnatal vocalizations of infant marmosets undergo a very similar developmental trajectory as infant human vocalizations: (1) they transition from noisy cries to tonal vocalizations that are adult-like (Takahashi et al., 2015); (2) these early vocalizations are produced in bouts that have a rhythmic pattern that changes over time (Zhang and Ghazanfar, 2016); and (3) vocal development is influenced by parental care and feedback (Takahashi et al., 2017; Gultekin and Hage, 2017; Gultekin and Hage, 2018). These across-species similarities in postnatal vocal development are also occurring at the same life history stage (though marmosets develop 12 times faster than humans [de Castro Leão et al., 2009]), supporting the idea that this species may also be a good model for prenatal human behaviors.

We tracked the fetal orofacial movements of marmoset monkeys that potentially represent those movements necessary for their neonatal contact-calling. These contact calls consist of both immature and mature-sounding versions (but mainly the former early in postnatal life) (Takahashi et al., 2015) and whose orofacial movements are indistinguishable from each other. Importantly, neonatal marmoset contact vocalizations consist of orofacial movements that are spatiotemporally distinct from any other vocalization in the marmoset’s vocal repertoire. They therefore provided a template: we could compare prenatal mouth movements with those produced during neonatal contact calling. We quantified and characterized the prenatal trajectory of these movements to address the following questions: if and how do mouth movements differentiate from more global bodily movements, and if and how do mouth movements differentiate prenatally into those used for neonatal vocalizations?

We performed non-invasive ultrasound imaging on awake, pregnant marmoset monkeys (n=4 pregnancies across two marmoset monkeys). Marmosets typically produce dizygotic twins at birth but not always (Harris et al., 2014). In our sample, the four pregnancies consisted of one singleton, two sets of twins and one set of quadruplets. Since individual fetuses cannot be routinely identified via ultrasound, each pregnancy with more than one fetus was treated as a composite of a single subject. To track developmental trajectories, we densely sampled fetal orofacial movements from the day the face was clearly discernible to the day before birth (~embryonic day [E]95–146; 2–3 times per week with imaging sessions lasting between 15 and 45 min; 14–17 sessions per pregnancy; a total of 64 sessions). We then compared these movements to signature features of infant vocalizations in the first week after birth (postnatal day [PD] 1–7; n=7). We were thus able to observe the developmental trajectory of orofacial movements necessary for vocal production from the fetal to infant stage.

We had two basic hypotheses. The first was that fetal orofacial movements would initially be linked to movements of other body parts. Ultrasound studies of human fetuses support this hypothesis, showing that isolated jaw movements appear late in development, weeks after the onset of spontaneous general movements involving many body parts (Fagard et al., 2018; Kurjak et al., 2004). Our second hypothesis was that signature patterns of articulatory movements related to postnatal marmoset contact calls would emerge prenatally. Investigations of the prenatal movements of rats, and prehatching movements of birds, show that seemingly unorganized action patterns are followed by organized action patterns that are reminiscent of postnatal or post-hatching movements. (Bekoff and Lau, 1980; Brumley and Robinson, 2010; Smotherman and Robinson, 1988; Bradley, 1999; Hamburger, 1963; Kuo, 1932; Provine, 1980).

In each imaging session, we specifically targeted the face as our region of interest, which allowed us to track both orofacial movements and head movements. From these videos, we did a frame-by-frame analysis (Figure 1A). The developmental change was readily noticeable, and ostensibly in keeping with our expectation that orofacial and head movements move together more often in the younger compared to the older fetus (Figure 1B–C; Videos 1 and 2).

Figure 1

Download asset Open asset

Video 1

Download asset

Video 2

Download asset

One of the hallmarks of development is increasing ‘order’ in behavior through self-organization. Since the relationship between fetal head and orofacial movements in marmosets changed over time, we tested if, for these two movement behaviors, there is concomitant increase of structure in their action patterns, by mapping the sequence of five different states: independent orofacial movements (State 1); independent head movements (State 2); orofacial movements followed by overlapping head movements (State 3); head movements followed by overlapping orofacial movements (State 4); and synchronous orofacial and head movements (State 5). The fetal behavioral states are seen to be more numerous and variable in the early weeks of gestation, when compared to late gestation. Figure 2A shows the proportion of states through gestation for the population data; Figure 2B shows the state diagrams for exemplar sessions selected from different gestational periods of one pregnancy.

Figure 2

Download asset Open asset

We quantified these state changes across gestational days in two ways. First, Shannon entropy was used to measure behavioral variability within each session (Figure 2C). For determining the developmental trajectory, the best polynomial-fit order using Akaike’s information criterion (AIC) was found to be three. Behavioral variability decreased through gestation and then remained steady at ~0.8 bits (the maximum entropy for a behavior with 5 possible states is 2.32 bits). The decrease in entropy (p<0.001, test of nullity of the relation between gestational day and entropy) indicates that as the fetus gets older, there is increasing movement structure.

Second, we performed a Kullback-Leibler divergence test to quantify the behavioral change through fetal life (Figure 2D). The average state distribution of E93-99 (n=6) was compared with the state distributions of all the imaging days. The best polynomial-fit order for the divergence estimates was one. The resulting linear fit with a positive slope (p<0.001) indicates that with increasing gestational age, fetal behavior—with respect to orofacial and head movements—becomes increasingly different from the first imaging day of orofacial movement. Thus, what we see is that in the young fetus, orofacial and head body parts move in a number of different ways—independently, and also often together in various combinations (orofacial-head, head-orofacial, or synchronous onset), but when close to birth, orofacial and head body parts are, with rare exception, moved separately. This suggests that these two motor regions are decoupled in the late-stage fetus, which allows the newly born infant to use head and orofacial movements adaptively; in marmoset infants, this would putatively be for the different purposes of orienting, feeding, and vocalizing.

To further quantify the developmental change, we examined the profiles of orofacial movements (n=1977) and head movements (n=1216) independently. We calculated the rates of orofacial and head movements (number of movements per hour) in each imaging session. The best polynomial-fit order was two for the orofacial movements, and one for head movements. The occurrence of orofacial movements showed an inverse U-shaped profile; steadily increasing from ~E93-105, hitting a peak from ~E106-131 and declining from ~E119-131 (Figure 2E, in red). The head movements, however, showed a linear trend. For both orofacial and head movements, the same patterns were observed in the individual pregnancies (Figure 2F; clockwise from top left: singleton, quadruplets, twins, and twins).

To study the putative ‘linked’ relationship between these two motor patterns, we calculated the percentage of temporal overlap between them (total instances of overlap divided by the total number of movements). A first order polynomial curve was found to be the best fit for determining the overlap profile. We found that the overlap of orofacial and head movements decreases over the course of fetal development (Figure 2G in purple, p<0.001). One possibility is that, since movements decrease in general toward the end of gestation, the decrease in overlapping movements simply becomes less probable. To account for this, we did a permutation test to be certain the decline in overlapping movements was independent of general movement decline. Orofacial and head movements within each session were independently shuffled, keeping the duration and latency distributions intact. Percent overlap was calculated for the permuted data, and a first order polynomial curve was fitted. The permutation test generated a flat profile (p=0.057 for the significance of the mean regression line) compared to the overlap decline seen in the actual data, confirming that the decline in overlap is not simply due to the decline in overall movement quantity (Figure 2G). Another possibility is that movement duration decreases throughout gestation, which would, therefore, decrease the probability of overlapping movements. This was not the case: the durations of mouth movements increased with time (p=0.024) and the durations of head movements remained steady throughout gestation (p=0.093). Thus, consistent with our prediction that mouth movements would be linked to other body parts, orofacial and head movements are coupled early in gestation but become increasingly independent. This held true for each of the four pregnancies (Figure 2H; clockwise from top left: singleton, quadruplets, twins, and twins).

Regarding vocalizations, we next focused only on the patterns of fetal orofacial movements to test the hypothesis that they gradually exhibited the temporal organization of the infant marmoset monkey’s rhythmic contact calling (Zhang and Ghazanfar, 2016). To do this, we quantified the temporal pattern of contact calling by neonatal marmosets (within ~24 hr of birth); these were the same individuals that were imaged as fetuses (n=7; one singleton, two sets of twins, and two surviving individuals from the quadruplets). Contact calling was induced by briefly separating infants from caregivers; we acquired ~5 min of audio recordings and ~5 min of both video and audio. In both media, infant contact calls are easily identifiable by their long duration and many syllables (Takahashi et al., 2015; Zhang and Ghazanfar, 2016; Figure 3A). We compared the recordings of infants producing contact calls to imaging data from fetuses producing orofacial movements (Figure 3B shows stills from the videos of an infant at P1 and a fetus at E118). The upper and lower jaws were tracked frame-by-frame to generate temporal profiles of the movements. We found that a subset of the fetal orofacial movements was a match to the infant contact call in terms of overall temporal profile, duration, and syllable number. (Figure 3C–F; Video 3).

Figure 3

Download asset Open asset

Video 3

Download asset

The cyclic nature of the infant contact call is fully captured in its temporal profile. We therefore matched the temporal profiles of fetal orofacial movements to those of infant contact calls. To do this, we took the video recordings (n=25) of neonatal marmoset infants producing contact calls and did a frame-by-frame tracking of the upper and lower jaw positions to generate contact call profiles. Multiple contact call profiles of the same syllable number were averaged and smoothed to create a set of contact call ‘templates’ (n=7; 1–7 syllables). To get the temporal profiles of fetal orofacial movements, we used a custom-made computer vision program that tracked each movement. Only temporal profiles that matched the manually calculated syllable number features of the movement were included in the analysis (n=414). We smoothed these profiles using a Savitzky-Golay smoothing filter (polynomial degree = 3; sliding window size = 9). We then used dynamic time warping (DTW) analysis (Keogh and Ratanamahatana, 2005) to measure the similarity between the temporal profiles of the fetal orofacial movements and the infant contact call templates. Every fetal orofacial movement profile was compared with every contact call template, and the one with the smallest DTW cost (i.e. the closest similarity) was chosen. The median DTW cost was calculated for each session (n=57). Our prediction was that the DTW cost would decrease through gestation, indicating that the temporal profiles of the fetal orofacial movements are getting closer to those of infant contact calls (Figure 4A). We found this to be the case (Figure 4B). A multiple linear regression analysis controlling for the different pregnancies shows that gestational age predicts the prevalence of ‘contact call-like’ temporal profiles (β±SE = –0.005±0.003, t=–1.97; F[1,56]=4.60; p=.037). In all four pregnancies, fetal orofacial movements increasingly match contact call temporal profiles as gestation progresses (Figure 4C; clockwise from top left: singleton, quadruplets, twins, and twins).

Figure 4

Download asset Open asset

We further quantified the developmental change leading up to this match in a number of ways. We first looked at the durations of fetal orofacial movements over the course of gestation (n=1845; Figure 4D). The median, 75th and 25th percentile durations were calculated for each session. To determine the developmental trajectory, the best polynomial-fit order using AIC was found to be three for the median values. The same polynomial order was used for generating curves for the 75th and 25th percentiles values. Kernel density estimation was used to visualize the duration density change over the course of gestation. The durations of orofacial movements show an increasing trend. Comparing the fetal orofacial movement durations to those of P1 infant contact calls, the fetal movements are seen to increasingly approach the P1 durations (Figure 4E; n=120; median duration = 3.69s; SE = 0.14).

Since contact calls typically contain multiple syllables, we calculated the ‘syllable number’ of the fetal orofacial movements (n=1845; Figure 4F). This is the number of individual orofacial movements separated by <500 ms (i.e. the same criteria used for counting actual contact call syllables in infants Zhang and Ghazanfar, 2016). The median, 75th and 25th percentile syllable numbers were calculated for each session. The best polynomial-fit order using AIC was found to be two for the median values. The same polynomial order was used to fit curves for the 75th and 25th percentiles values. The polynomial curves and kernel density estimation show that the syllable numbers of fetal orofacial movements increase through gestation. Fetal syllable numbers get closer to the syllable numbers of contact calls produced on P1 (Figure 4G; n=120; median = 5; SE = 0.18).

To directly compare the features of fetal orofacial movements with signature features of the infant contact call, we established criteria for identifying the contact call based on duration and syllable number. Contact calls tend to have long durations, and multiple syllables when compared to other infant calls. An infant call within the duration range of 3.69–6.5 s and syllable number range 5–9 has a 97% likelihood of being a contact call. To track the development of ‘contact call-like’ fetal orofacial movements, we calculated, for each session, the percentage of movements that matched the duration and syllable number features of P1 contact calls (Figure 4H). A multiple linear regression analysis controlling for the different pregnancies shows that fetal orofacial movements increasingly match the contact call profile as the fetus gets older (β±SE = 0.15±0.04, t=3.44; F(1, 63)=12.09, p=0.001). All four pregnancies were seen to have more ‘contact call-like’ movements over the course of gestation (Figure 4I; clockwise from top left: singleton, quadruplets, twins, twins). (Statistics for each individual pregnancies are available in Supplementary file 1).

To control for the possibility that the fetal orofacial movement profile is a generic one and not specifically linked to contact calls, we compared it to the two other orofacial movements common in the first weeks of postnatal life: twitter vocalizations and licking. When separated from caregivers, marmoset infants also produce bouts of another type of vocalization known as twitters. Relative to contact calls, twitters are short duration calls with fewer syllables. First, we checked if this contrasting type of vocalization produced by the P1 infants (n=109 twitters) follows a similar developmental trajectory. An infant call within the duration range of 0.96–1.5 s and syllable number range 2–3 has a 86% likelihood of being a twitter call (and a 2% likelihood of being a contact call). We then calculated, for each session, the percentage of fetal movements that matched the duration and syllable number features of P1 twitter calls (Figure 5A). A multiple linear regression analysis controlling for the different pregnancies reveals that—in contrast to what is seen with the contact call—fetal orofacial movements do not increasingly match the twitter profile as the fetus gets older (β±SE = 0.06±0.06, t=0.99; F(1, 63)=0.87, p=0.35).

Figure 5

Download asset Open asset

We observed that when separated from their caregivers the infants often produced licking movements. (Note: weaning begins at ~30 days in marmosets, therefore chewing movements are not seen in neonates.) We measured the duration and syllable number features of these movements in the same infants, from ages P1-7 (n=37) and found that licks have greater variability than the contact call both in duration (Figure 5B; median = 1.6s; SE = 0.83) and ‘syllable’ number (number of jaw movements) (Figure 5C; median = 2; SE = 0.85). We found a 0% likelihood that a movement within the duration range of 3.69–6.5 s and syllable number range of 5–9 (our contact call criteria) is a licking-related movement (Figure 5D). Infant licking movements are distinct from their contact calls.

We then tracked the development of ‘lick-like’ orofacial movements in the fetus. The licks of P1-7 infants tend to be short with few syllables, or very long with a high number of syllables. To capture the maximum possible infant licks, we established separate criteria for each of these types of licks: duration range of 0.33–3.77 s and syllable number range of 1–6 for the short licks and duration range of 6.67–14.29 s and syllable number range of 10–16 for the long licks. Together they capture 83.78% of the infant licks (and 0% of the infant contact calls). We then calculated, for each session, the percentage of fetal movements that matched the features of licks (Figure 5E). We identified more ‘lick-like’ orofacial movements than ‘contact call-like’ movements in the fetus, but a multiple linear regression analysis controlling for the different pregnancies reveals that fetal orofacial movements do not increasingly match the lick profile as the fetus gets older (β±SE = 0.10±0.11, t=0.97; F(1, 63)=0.78, p=0.38). These comparisons establish that the changes in movements we have tracked are specifically related to the rhythmic contact calls made by neonatal marmosets. It is worth noting that an increase in lick-like movements over the course of development could have been missed due to ceiling effects, i.e., they may have developed at an earlier time in gestation. Nevertheless, we can conclude is that in the gestation period under observation (~E93 to birth), there is no significant increase in lick-like movements. However, as we hypothesized, the marmoset contact call production has a period of prenatal development.

The neonate’s well-being hinges on using vocalizations to solicit attention. In human and non-human primates, these neonatal vocalizations transform into mature vocal signals through continuous and reciprocal interactions among the infant’s advancing vocal apparatus, neural circuitry, and interactions with caregivers (Ghazanfar and Liao, 2018; Teramoto et al., 2017; Thelen et al., 1991). Our study focused on the prenatal origins of those very first vocalizations; particularly the development of their spatiotemporal signatures, which we tracked using vocalization-related mouth movements. Using fetal marmoset monkeys, we show that early mouth movements are tightly coupled to head movements. As gestation progresses, the movements decouple, ultimately leading to the formation of two independent movements. In parallel, a subset of fetal orofacial movements gradually assumes the pattern necessary to produce the immature-sounding contact calls neonates use to solicit care (Huang et al., 2020) and the parental interactions used to ratchet their vocal development (Takahashi et al., 2015; Takahashi et al., 2017; Gultekin and Hage, 2017; Gultekin and Hage, 2018; Takahashi et al., 2016). Aspects of the sensory-motor development necessary for vocalizing are occurring prenatally, even before the production of sound.

That body parts (head and mouth movements) were initially linked in marmoset fetuses is consistent with the developmental pattern of fetal limb movements across many species (Hamburger, 1963; Barron, 1941; Coghill, 1929; Oppenheim, 1974; Robinson, 1988). For example, early in gestation, embryonic chicks move their limbs and wings together. It is only in later prehatching life that these two body parts uncouple in their movements and can perform separate functions (Bradley, 1999). In sheep, the first limb movements are paired with head and neck motions, but these different body parts acquire increasing autonomy over the course of development (Barron, 1941). These fetal limb movements play an important role in the formation and organization of the spinal cord and CNS (Granmo et al., 2008). However, the development of these head and mouth action patterns in marmosets involve central pattern generators located in the brainstem not the spinal cord (Barlow et al., 2010; Moore et al., 2014). This suggests that the principles of fetal development—whereby linked movements involving multiple body parts are gradually unlinked—are general.

One question is whether the underlying mechanisms are the same for the patterns of change seen for head-mouth movement differentiation versus the differentiation between limb movements and other parts of the body. Studies of chicks suggest that the constraints of the egg help differentiate motor units and construct species-typical patterns of embryonic motility (Bekoff, 2001). Increasing uterine constraint has also been used to explain the common inversed-shaped pattern of limb movement frequency observed in many species (Barron, 1941; Oppenheim, 1974; Robinson, 1988; Gottlieb, 1976; Roodenburg et al., 1991). In our study of marmosets and in humans (Roodenburg et al., 1991), the frequency of mouth movements over the course of gestation also showed an inversed U-shaped trajectory. Do they decrease in number because, as the fetus grows larger, there is less room to make orofacial movements? This seems unlikely in our study. In the four pregnancies we monitored, the number of fetuses (and thus the space constraints) varied, but there was no indication that there was any related difference in the developmental trajectory of mouth movements. Consistent with this idea, a study of fetal rats compared their activity levels under three different, decreasing space constraints: the uterine environment, outside the uterus but with extra-embryonic membranes intact, and with all membranes removed (Smotherman and Robinson, 1988). In all cases, activity profiles still exhibited the familiar inversed U-shaped profile (Smotherman and Robinson, 1988). These data suggest that the shape of the developmental trajectory is influenced by neural changes, possibly from the developing forebrain (Barron, 1941; Decker and Hamburger, 1967). Indeed, a computational model of motor development supports the idea that spontaneous motor activity can self-organize into reflex actions, and that these actions are subsequently modulated to produced coordinated, adaptive behaviors (Marques et al., 2014).

In utero interactions between the species-typical body and brain, and between the body and its environment are hypothesized to drive the development of muscle and neural organization, and, ultimately, lead to the development of species-typical behavior (Bekoff, 2001). This idea is supported by computer simulations of fetal development that model fetal motor behavior and spinal circuit changes of different mammalian species within a uterine environment (Gottlieb, 1976; Roodenburg et al., 1991; Decker and Hamburger, 1967; Marques et al., 2014). At beginning of simulations, all fetuses had different musculoskeletal bodies but the same undifferentiated neural circuits. Following spontaneous activity, the simulated fetuses developed species typical motor patterns, e.g., the human fetus developed alternating leg movements and the zebrafish embryo developed coordinated side-to-side movements. A by-product of these simulations was that individual muscle contractions were initially part of larger undifferentiated motor units (e.g. a single unit controlling the head and the trunk), but as development progressed, these large muscle units subdivided into smaller, more precisely controlled units (e.g. separate control of the head and the trunk). Our finding that head and mouth motor units are linked early in gestation and uncouple with age provide additional support for this hypothesis.

A subset of fetal marmoset mouth movements were precursors to the movements needed to produce contact calls as neonates; over gestation, they became increasingly similar in their temporal structure to neonatal articulatory movements. Likewise, the movement patterns of fetal rodent limbs are seen to get quantitatively and qualitatively closer to those of the neonate (Brumley and Robinson, 2010; Kleven et al., 2004; Robinson and Kleven GA, 2005). These same phenomena are observed in chicken embryos (Bradley and Sebelski, 2000; Sharp et al., 1999). These studies in rodents and birds suggest that experiences in-utero play a key role in establishing early postnatal movements. For example, fetal rats modify their limb movements in response to physical constraints, exhibiting motor learning in utero (Robinson and Kleven GA, 2005). The simulations described above also support the notion that in utero learning via sensory feedback generated by fetal movements lead to species-typical stepping or swimming actions (Kuniyoshi, 2019; Mori and Kuniyoshi, 2010; Yamada et al., 2010; Yamada and Kuniyoshi, 2012; Yamada et al., 2016; Mori, 2012). Thus, it does not seem too far-fetched to suggest that self-generated experience leads to the convergence of the temporal structure of marmoset fetal mouth movements toward the pattern of neonatal vocal articulatory movements (at least for contact calls). Self-generated learning of this type would be of the same kind that is observed postnatally in mammals, whereby myoclonic twitches during sleep help construct sensorimotor maps in the brain (Dooley and Blumberg, 2018; Dooley et al., 2020; Blumberg et al., 2013a)(for review, see Blumberg et al., 2013b).

Here are some caveats to our study. Our study focused only on contact calls because they are among the first vocalizations produced by infants (albeit immature sounding) and are important for soliciting both vocal responses (Takahashi et al., 2016) and retrieval from caregivers (Huang et al., 2020). The most obvious parallel to marmoset monkey contact calling (and likely all infant contact calling in mammals and birds) is crying by human infants. The comparison, however, is not so simple. The contact vocalizations that infant marmosets use to solicit caregiver attention are also used by the adults to maintain social contact with conspecifics; this type of vocal behavior seems to be true for all nonhuman primates (Newman, 2004). In humans, communication with caregivers through crying soon gives way to other types of vocalizations and eventually speech (Oller, 2000). Next, in pregnancies with multiple fetuses, we treated them as a single composite subject. This was because there was no way to identify the same individuals across days. Two findings mitigate against this issue: all four pregnancies exhibited the same behavioral profiles and developmental trajectories, and the early postnatal vocal output of twins are nearly identical in their patterns of output (more so than in comparison to their non-twin siblings), seemingly driven by arousal levels regulated by their caregivers (Zhang and Ghazanfar, 2016).

Another caveat is that while we implicitly interpreted the fetal orofacial movements we observed as being ‘spontaneous’ as other similar studies have, it is possible that these movements could have been initiated by activation of the developing sensory system (Hamburger, 1963). However, we did not observe any correlation with, e.g., application of the ultrasound probe and fetal movements, and a pilot study by us to elicit fetal movements via sound playback near the mother’s belly was unsuccessful. Relatedly, our previous work (Zhang and Ghazanfar, 2016) showed the initiation and duration of infant vocal output is linked to a 0.1 Hz autonomic rhythm and that the temporal structure of those sequences are correlated with the respiratory cycle. Thus, one possibility is that the timing of fetal bodily and/or orofacial movements are also tied to such cycles, or at least to the autonomic rhythm (in humans at least, fetal ‘practice’ breathing movements are intermittent). Early on in our study, we attempted to investigate this link but we could not simultaneously monitor heart rate (in mothers or fetuses) while also monitoring orofacial movements. Finally, our sample size consisted of four pregnancies across two unrelated mothers. The imaging and analysis were very time intensive, and we opted to longitudinally sample individual pregnancies very densely as opposed to increasing the number of subjects. This allowed us to better map the shape of developmental trajectories (Adolph et al., 2008).

Marmoset monkeys offer a special opportunity for comparative investigations of vocal development with humans. Like humans, marmosets exhibit a high degree of prosociality (Burkart et al., 2014) (and related phenotypes linked to ‘self-domestication’ [Ghazanfar et al., 2020]), show context-dependent vocal control and plasticity in the timing, intensity and spectral features of their vocalizations (Choi et al., 2015; Takahashi et al., 2013; Zürcher et al., 2019; Liao et al., 2018), and infant marmosets undergo a period of vocal learning that is influenced by parents (Takahashi et al., 2015; Takahashi et al., 2017; Gultekin and Hage, 2017; Gultekin and Hage, 2018). All of these similarities may be related to the fact that both humans and marmosets are cooperative breeders and are born altricial relative to other primates (Varella and Ghazanfar, 2021). Thus, we predict that the prenatal patterns of vocalization-related orofacial movements in marmosets we observed here is of the same kind in humans. This would open up possibilities to use marmoset monkeys to investigate how very early prenatal orofacial motor behaviors can be diagnostic of later vocal behaviors and prelinguistic vocal learning.

All data generated or analysed during this study are available on DRYAD https://doi.org/10.5061/dryad.m905qfv1x.

1. 1. Ghazanfar AA

   (2022) Dryad Digital Repository

   Data from: Prenatal development of neonatal vocalizations.

   https://doi.org/10.5061/dryad.m905qfv1x

1. 1. Adolph KE
   2. Robinson SR
   3. Young JW
   4. Gill-Alvarez F

   (2008) What is the shape of developmental change?

   Psychological Review 115:527–543.

   https://doi.org/10.1037/0033-295X.115.3.527

   • PubMed
   • Google Scholar
2. 1. Akaike H

   (1981)

   A new look at the statistical-model identification

   Current/Contents/Engineering Technology & Applied Sciences 51:e2.

   • Google Scholar
3. Book

   1. Barlow SM
   2. Radder JPL
   3. Radder ME
   4. Radder AK

   (2010) Central pattern generators for orofacial movements and speech

   In: Muller CP, Cunningham KA, editors. Handbook of Behavioral Neuroscience. ScienceDirect. pp. 351–369.

   https://doi.org/10.1016/B978-0-12-374593-4.00033-4

   • Google Scholar
4. 1. Barron DH

   (1941) The functional development of some mammalian neuromuscular mechanisms

   Biological Reviews 16:1–33.

   https://doi.org/10.1111/j.1469-185X.1941.tb01093.x

   • Google Scholar
5. 1. Bekoff A
   2. Lau B

   (1980) Interlimb coordination in 20-day-old rat fetuses

   The Journal of Experimental Zoology 214:173–175.

   https://doi.org/10.1002/jez.1402140207

   • PubMed
   • Google Scholar
6. 1. Bekoff A

   (2001) Spontaneous embryonic motility: an enduring legacy

   International Journal of Developmental Neuroscience 19:155–160.

   https://doi.org/10.1016/s0736-5748(00)00089-7

   • PubMed
   • Google Scholar
7. 1. Bergman TJ
   2. Beehner JC
   3. Painter MC
   4. Gustison ML

   (2019) The speech-like properties of nonhuman primate vocalizations

   Animal Behaviour 151:229–237.

   https://doi.org/10.1016/j.anbehav.2019.02.015

   • Google Scholar
8. 1. Blumberg MS
   2. Coleman CM
   3. Gerth AI
   4. McMurray B

   (2013a) Spatiotemporal structure of REM sleep twitching reveals developmental origins of motor synergies

   Current Biology 23:2100–2109.

   https://doi.org/10.1016/j.cub.2013.08.055

   • PubMed
   • Google Scholar
9. 1. Blumberg MS
   2. Marques HG
   3. Iida F

   (2013b) Twitching in sensorimotor development from sleeping rats to robots

   Current Biology 23:R532–R57.

   https://doi.org/10.1016/j.cub.2013.04.075

   • PubMed
   • Google Scholar
10. 1. Bradley NS

    (1999) Transformations in embryonic motility in chick: kinematic correlates of type I and II motility at E9 and E12

    Journal of Neurophysiology 81:1486–1494.

    https://doi.org/10.1152/jn.1999.81.4.1486

    • PubMed
    • Google Scholar
11. 1. Bradley NS
    2. Sebelski C

    (2000) Ankle restraint modifies motility at E12 in chick embryos

    Journal of Neurophysiology 83:431–440.

    https://doi.org/10.1152/jn.2000.83.1.431

    • PubMed
    • Google Scholar
12. Book

    1. Brumley MR
    2. Robinson SR

    (2010) Experience in the perinatal development of action systems

    In: Lidz JL, Snyder W, Pate J, editors. Oxford Handbook of Developmental. Oxford University Press. pp. 181–209.

    https://doi.org/10.1093/oxfordhb/9780195314731.013.0010

    • Google Scholar
13. 1. Burkart JM
    2. Allon O
    3. Amici F
    4. Fichtel C
    5. Finkenwirth C
    6. Heschl A
    7. Huber J
    8. Isler K
    9. Kosonen ZK
    10. Martins E
    11. Meulman EJ
    12. Richiger R
    13. Rueth K
    14. Spillmann B
    15. Wiesendanger S
    16. van Schaik CP

    (2014) The evolutionary origin of human hyper-cooperation

    Nature Communications 5:1–9.

    https://doi.org/10.1038/ncomms5747

    • PubMed
    • Google Scholar
14. 1. Chandrasekaran C
    2. Trubanova A
    3. Stillittano S
    4. Caplier A
    5. Ghazanfar AA

    (2009) The natural statistics of audiovisual speech

    PLOS Computational Biology 5:e1000436.

    https://doi.org/10.1371/journal.pcbi.1000436

    • PubMed
    • Google Scholar
15. 1. Choi JY
    2. Takahashi DY
    3. Ghazanfar AA

    (2015) Cooperative vocal control in marmoset monkeys via vocal feedback

    Journal of Neurophysiology 114:274–283.

    https://doi.org/10.1152/jn.00228.2015

    • PubMed
    • Google Scholar
16. Book

    1. Coghill GE

    (1929)

    Anatomy and the Problem of Behavior

    New York: MacMillan Co.

    • Google Scholar
17. 1. de Castro Leão A
    2. Duarte Dória Neto A
    3. de Sousa MBC

    (2009) New developmental stages for common marmosets (Callithrix jacchus) using mass and age variables obtained by K-means algorithm and self-organizing maps (SOM

    Computers in Biology and Medicine 39:853–859.

    https://doi.org/10.1016/j.compbiomed.2009.05.009

    • PubMed
    • Google Scholar
18. 1. Decker JD
    2. Hamburger V

    (1967) The influence of different brain regions on periodic motility of the chick embryo

    The Journal of Experimental Zoology 165:371–383.

    https://doi.org/10.1002/jez.1401650306

    • PubMed
    • Google Scholar
19. 1. Dooley JC
    2. Blumberg MS

    (2018) Developmental “awakening” of primary motor cortex to the sensory consequences of movement

    eLife 7:e41841.

    https://doi.org/10.7554/eLife.41841

    • PubMed
    • Google Scholar
20. 1. Dooley JC
    2. Glanz RM
    3. Sokoloff G
    4. Blumberg MS

    (2020) Self-generated whisker movements drive state-dependent sensory input to developing barrel cortex

    Current Biology 30:2404–2410.

    https://doi.org/10.1016/j.cub.2020.04.045

    • PubMed
    • Google Scholar
21. 1. Fagard J
    2. Esseily R
    3. Jacquey L
    4. O’Regan K
    5. Somogyi E

    (2018) Fetal origin of sensorimotor behavior

    Frontiers in Neurorobotics 12:23.

    https://doi.org/10.3389/fnbot.2018.00023

    • PubMed
    • Google Scholar
22. 1. Ghazanfar AA
    2. Takahashi DY

    (2014) The evolution of speech: vision, rhythm, cooperation

    Trends in Cognitive Sciences 18:543–553.

    https://doi.org/10.1016/j.tics.2014.06.004

    • PubMed
    • Google Scholar
23. 1. Ghazanfar AA
    2. Liao DA

    (2018) Constraints and flexibility during vocal development: Insights from marmoset monkeys

    Current Opinion in Behavioral Sciences 21:27–32.

    https://doi.org/10.1016/j.cobeha.2017.11.015

    • PubMed
    • Google Scholar
24. 1. Ghazanfar AA
    2. Kelly LM
    3. Takahashi DY
    4. Winters S
    5. Terrett R
    6. Higham JP

    (2020) Domestication phenotype linked to vocal behavior in marmoset monkeys

    Current Biology 30:5026–5032.

    https://doi.org/10.1016/j.cub.2020.09.049

    • PubMed
    • Google Scholar
25. 1. Gingras JL
    2. Mitchell EA
    3. Grattan KE

    (2005) Fetal homologue of infant crying

    Archives of Disease in Childhood. Fetal and Neonatal Edition 90:F415–F418.

    https://doi.org/10.1136/adc.2004.062257

    • PubMed
    • Google Scholar
26. 1. Goldstein MH
    2. Schwade JA

    (2008) Social feedback to infants’ babbling facilitates rapid phonological learning

    Psychological Science 19:515–523.

    https://doi.org/10.1111/j.1467-9280.2008.02117.x

    • PubMed
    • Google Scholar
27. 1. Gottlieb G

    (1976) Conceptions of prenatal development: behavioral embryology

    Psychological Review 83:215–234.

    https://doi.org/10.1037//0033-295X.83.3.215

    • PubMed
    • Google Scholar
28. 1. Granmo M
    2. Petersson P
    3. Schouenborg J

    (2008) Action-based body maps in the spinal cord emerge from a transitory floating organization

    The Journal of Neuroscience 28:5494–5503.

    https://doi.org/10.1523/JNEUROSCI.0651-08.2008

    • PubMed
    • Google Scholar
29. 1. Gultekin YB
    2. Hage SR

    (2017) Limiting parental feedback disrupts vocal development in marmoset monkeys

    Nature Communications 8:14046.

    https://doi.org/10.1038/ncomms14046

    • PubMed
    • Google Scholar
30. 1. Gultekin YB
    2. Hage SR

    (2018) Limiting parental interaction during vocal development affects acoustic call structure in marmoset monkeys

    Science Advances 4:eaar4012.

    https://doi.org/10.1126/sciadv.aar4012

    • PubMed
    • Google Scholar
31. 1. Hamburger V

    (1963) Some aspects of the embryology of behavior

    The Quarterly Review of Biology 38:342–365.

    https://doi.org/10.1086/403941

    • PubMed
    • Google Scholar
32. 1. Harris RA
    2. Tardif SD
    3. Vinar T
    4. Wildman DE
    5. Rutherford JN
    6. Rogers J
    7. Worley KC
    8. Aagaard KM

    (2014) Evolutionary genetics and implications of small size and twinning in callitrichine primates

    PNAS 111:1467–1472.

    https://doi.org/10.1073/pnas.1316037111

    • PubMed
    • Google Scholar
33. 1. Hopkins B
    2. von Wulfften Palthe T

    (1987) The development of the crying state during early infancy

    Developmental Psychobiology 20:165–175.

    https://doi.org/10.1002/dev.420200206

    • PubMed
    • Google Scholar
34. 1. Huang J
    2. Cheng X
    3. Zhang S
    4. Chang L
    5. Li X
    6. Liang Z
    7. Gong N

    (2020) Having infants in the family group promotes altruistic behavior of marmoset monkeys

    Current Biology 30:4047–4055.

    https://doi.org/10.1016/j.cub.2020.07.045

    • PubMed
    • Google Scholar
35. 1. Jaquish CE
    2. Toal RL
    3. Tardif SD
    4. Carson RL

    (1995) Use of ultrasound to monitor prenatal growth and development in the common marmoset (Callithrix jacchus

    American Journal of Primatology 36:259–275.

    https://doi.org/10.1002/ajp.1350360402

    • PubMed
    • Google Scholar
36. 1. Keogh E
    2. Ratanamahatana CA

    (2005) Exact indexing of dynamic time warping

    Knowledge and Information Systems 7:358–386.

    https://doi.org/10.1007/s10115-004-0154-9

    • Google Scholar
37. 1. Kleven GA
    2. Lane MS
    3. Robinson SR

    (2004) Development of interlimb movement synchrony in the rat fetus

    Behavioral Neuroscience 118:835–844.

    https://doi.org/10.1037/0735-7044.118.4.835

    • PubMed
    • Google Scholar
38. 1. Kuniyoshi Y

    (2019) Fusing autonomy and sociality via embodied emergence and development of behaviour and cognition from fetal period

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 374:1771.

    https://doi.org/10.1098/rstb.2018.0031

    • PubMed
    • Google Scholar
39. 1. Kuo ZY

    (1932) Ontogeny of embryonic behavior in aves. IV. The influence of embryonic movements upon the behavior after hatching

    Journal of Comparative Psychology 14:109–122.

    https://doi.org/10.1037/h0071451

    • Google Scholar
40. 1. Kurjak A
    2. Stanojevic M
    3. Andonotopo W
    4. Salihagic-Kadic A
    5. Carrera JM
    6. Azumendi G

    (2004) Behavioral pattern continuity from prenatal to postnatal life--a study by four-dimensional (4D) ultrasonography

    Journal of Perinatal Medicine 32:346–353.

    https://doi.org/10.1515/JPM.2004.065

    • PubMed
    • Google Scholar
41. 1. Lameira AR
    2. Hardus ME
    3. Bartlett AM
    4. Shumaker RW
    5. Wich SA
    6. Menken SBJ

    (2015) Speech-like rhythm in a voiced and voiceless orangutan call

    PLOS ONE 10:e116136.

    https://doi.org/10.1371/journal.pone.0116136

    • PubMed
    • Google Scholar
42. 1. Liao DA
    2. Zhang YS
    3. Cai LX
    4. Ghazanfar AA

    (2018) Internal states and extrinsic factors both determine monkey vocal production

    PNAS 115:3978–3983.

    https://doi.org/10.1073/pnas.1722426115

    • PubMed
    • Google Scholar
43. 1. Mampe B
    2. Friederici AD
    3. Christophe A
    4. Wermke K

    (2009) Newborns’ cry melody is shaped by their native language

    Current Biology 19:1994–1997.

    https://doi.org/10.1016/j.cub.2009.09.064

    • PubMed
    • Google Scholar
44. 1. Marques HG
    2. Bharadwaj A
    3. Iida F

    (2014) From spontaneous motor activity to coordinated behaviour: a developmental model

    PLOS Computational Biology 10:e1003653.

    https://doi.org/10.1371/journal.pcbi.1003653

    • PubMed
    • Google Scholar
45. 1. Moore JD
    2. Kleinfeld D
    3. Wang F

    (2014) How the brainstem controls orofacial behaviors comprised of rhythmic actions

    Trends in Neurosciences 37:370–380.

    https://doi.org/10.1016/j.tins.2014.05.001

    • PubMed
    • Google Scholar
46. Conference

    1. Mori H
    2. Kuniyoshi Y

    (2010) A human fetus development simulation: Self-organization of behaviors through tactile sensation

    2010 IEEE 9th International Conference on Development and Learning. pp. 82–87.

    https://doi.org/10.1109/DEVLRN.2010.5578860

    • Google Scholar
47. Conference

    1. Mori H

    (2012)

    Is the developmental order of fetal behaviors self-organized in an uterine environment?

    2012 IEEE International Conference on Development and Learning and Epigenetic Robotics (ICDL. pp. 1–2.

    • Google Scholar
48. Book

    1. Newman JD

    (2004)

    The primate isolation call: a comparison with precocial birds and non-primate mammals

    In: Rogers LJ, Kaplan G, editors. Comparative Vertebrate Cognition. Springer New York, NY. pp. 171–187.

    • Google Scholar
49. Book

    1. Oller DK

    (2000) The Emergence of the Speech Capacity

    Psychology Press.

    https://doi.org/10.4324/9781410602565

    • Google Scholar
50. Book

    1. Oppenheim RW

    (1974)

    The ontogeny of behavior in the chick embryo

    In: Lehrman D, Rosenblatt J, editors. Advances in the Study of Behavior Vol 5. Academic Press. pp. 133–172.

    • Google Scholar
51. 1. Petitjean F
    2. Ketterlin A
    3. Gançarski P

    (2011) A global averaging method for dynamic time warping, with applications to clustering

    Pattern Recognition 44:678–693.

    https://doi.org/10.1016/j.patcog.2010.09.013

    • Google Scholar
52. 1. Provine RR

    (1980) Development of between-limb movement synchronization in the chick embryo

    Developmental Psychobiology 13:151–163.

    https://doi.org/10.1002/dev.420130207

    • PubMed
    • Google Scholar
53. 1. Risueno-Segovia C
    2. Hage SR

    (2020) Theta synchronization of phonatory and articulatory systems in marmoset monkey vocal production

    Current Biology 30:4276–4283.

    https://doi.org/10.1016/j.cub.2020.08.019

    • PubMed
    • Google Scholar
54. Book

    1. Robinson R

    (1988)

    Dimensions of Fetal Investigation

    In: Smotherman WP, Robinson R, editors. Behavior of the Fetus. The Teleford Press, INC. pp. 19–34.

    • Google Scholar
55. Book

    1. Robinson SR
    2. Kleven GA GA

    (2005)

    Learning to move before birth

    In: Hopkins B, Johnson SP, editors. Perinatal Development of Action Systems in the Rat. Praeger. pp. 131–175.

    • Google Scholar
56. 1. Roodenburg PJ
    2. Wladimiroff JW
    3. van Es A
    4. Prechtl HF

    (1991) Classification and quantitative aspects of fetal movements during the second half of normal pregnancy

    Early Human Development 25:19–35.

    https://doi.org/10.1016/0378-3782(91)90203-f

    • PubMed
    • Google Scholar
57. 1. Sharp AA
    2. Ma E
    3. Bekoff A

    (1999) Developmental changes in leg coordination of the chick at embryonic days 9, 11, and 13: uncoupling of ankle movements

    Journal of Neurophysiology 82:2406–2414.

    https://doi.org/10.1152/jn.1999.82.5.2406

    • PubMed
    • Google Scholar
58. Book

    1. Smotherman WP
    2. Robinson SR

    (1988) The uterus as environment. The ecology of fetal behaviour

    In: Blass EM, editors. Handbook of Behavioral Neurobiology. Plenum. pp. 149–196.

    https://doi.org/10.1007/978-1-4684-5421-5_5

    • Google Scholar
59. 1. Takahashi DY
    2. Narayanan DZ
    3. Ghazanfar AA

    (2013) Coupled oscillator dynamics of vocal turn-taking in monkeys

    Current Biology 23:2162–2168.

    https://doi.org/10.1016/j.cub.2013.09.005

    • PubMed
    • Google Scholar
60. 1. Takahashi DY
    2. Fenley AR
    3. Teramoto Y
    4. Narayanan DZ
    5. Borjon JI
    6. Holmes P
    7. Ghazanfar AA

    (2015) Language development the developmental dynamics of marmoset monkey vocal production

    Science 349:734–738.

    https://doi.org/10.1126/science.aab1058

    • PubMed
    • Google Scholar
61. 1. Takahashi DY
    2. Fenley AR
    3. Ghazanfar AA

    (2016) Early development of turn-taking with parents shapes vocal acoustics in infant marmoset monkeys

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 371:1693.

    https://doi.org/10.1098/rstb.2015.0370

    • PubMed
    • Google Scholar
62. 1. Takahashi DY
    2. Liao DA
    3. Ghazanfar AA

    (2017) Vocal learning via social reinforcement by infant marmoset monkeys

    Current Biology 27:1844–1852.

    https://doi.org/10.1016/j.cub.2017.05.004

    • PubMed
    • Google Scholar
63. 1. Teramoto Y
    2. Takahashi DY
    3. Holmes P
    4. Ghazanfar AA

    (2017) Vocal development in a Waddington landscape

    eLife 6:e20782.

    https://doi.org/10.7554/eLife.20782

    • PubMed
    • Google Scholar
64. Book

    1. Thelen E
    2. Krasnegor NA
    3. Rumbaugh DM
    4. Schiefelbusch RL
    5. Studdert-Kennedy M

    (1991)

    Biological and Behavioral Determinants of Language Development

    New Jersey: Lawrence Erlbaum Associates.

    • Google Scholar
65. 1. Varella TT
    2. Ghazanfar AA

    (2021) Cooperative care and the evolution of the prelinguistic vocal learning

    Developmental Psychobiology 63:1583–1588.

    https://doi.org/10.1002/dev.22108

    • PubMed
    • Google Scholar
66. 1. Warlaumont AS
    2. Richards JA
    3. Gilkerson J
    4. Oller DK

    (2014) A social feedback loop for speech development and its reduction in autism

    Psychological Science 25:1314–1324.

    https://doi.org/10.1177/0956797614531023

    • PubMed
    • Google Scholar
67. 1. Wolff PH

    (1969)

    The natural history of crying and other vocalizations in early infancy

    Determinants of Infant Behavior 1:81–111.

    • Google Scholar
68. Conference

    1. Yamada Y
    2. Mori H
    3. Kuniyoshi Y

    (2010)

    A Fetus and Infant Developmental Scenario: Self-organization of Goal-directed Behaviors Based on Sensory Constraints

    Proceedings of the Tenth International Conference on Epigenetic Robotics: Modeling Cognitive Development in Robotic Systems.

    • Google Scholar
69. Conference

    1. Yamada Y
    2. Kuniyoshi Y

    (2012) Embodiment guides motor and spinal circuit development in vertebrate embryo and fetus

    2012 IEEE International Conference on Development and Learning and Epigenetic Robotics (ICDL). pp. 1–6.

    https://doi.org/10.1109/DevLrn.2012.6400578

    • Google Scholar
70. 1. Yamada Y
    2. Kanazawa H
    3. Iwasaki S
    4. Tsukahara Y
    5. Iwata O
    6. Yamada S
    7. Kuniyoshi Y

    (2016) An embodied brain model of the human foetus

    Scientific Reports 6:27893.

    https://doi.org/10.1038/srep27893

    • PubMed
    • Google Scholar
71. 1. Zhang YS
    2. Ghazanfar AA

    (2016) Perinatally influenced autonomic system fluctuations drive infant vocal sequences

    Current Biology 26:1249–1260.

    https://doi.org/10.1016/j.cub.2016.03.023

    • PubMed
    • Google Scholar
72. 1. Zürcher Y
    2. Willems EP
    3. Burkart JM

    (2019) Are dialects socially learned in marmoset monkeys? Evidence from translocation experiments

    PLOS ONE 14:e0222486.

    https://doi.org/10.1371/journal.pone.0222486

    • PubMed
    • Google Scholar

Author details

1. Darshana Z Narayanan

   1. Princeton Neuroscience Institute, Princeton University, Princeton, United States
   2. Department of Psychology , Princeton University, Princeton, United States

   Contribution

   Conceptualization, Formal analysis, Investigation, Visualization, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3553-1875

2. Daniel Y Takahashi

   1. Princeton Neuroscience Institute, Princeton University, Princeton, United States
   2. Department of Psychology , Princeton University, Princeton, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Writing – review and editing

   For correspondence

   takahashiyd@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4972-001X

3. Lauren M Kelly

   1. Princeton Neuroscience Institute, Princeton University, Princeton, United States
   2. Department of Psychology , Princeton University, Princeton, United States

   Contribution

   Formal analysis, Writing – review and editing

   Competing interests

   No competing interests declared

4. Sabina I Hlavaty

   Department of Ecology and Evolutionary Biology, Princeton University, Princeton, United States

   Contribution

   Formal analysis, Investigation

   Competing interests

   No competing interests declared

5. Junzhou Huang

   Department of Computer Science and Engineering, The University of Texas at Arlington, Arlington, United States

   Contribution

   Software, Methodology

   Competing interests

   No competing interests declared

6. Asif A Ghazanfar

   1. Princeton Neuroscience Institute, Princeton University, Princeton, United States
   2. Department of Psychology , Princeton University, Princeton, United States

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing – original draft, Writing – review and editing

   For correspondence

   asifg@princeton.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1960-7470

• 2,270

  views

• 464

  downloads

• 4

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.