Sound iconicity in the transition from protosign to protospeech

1 Introduction

The question of whether human language originated from vocalizations, gestures, or a combination of modalities has been the source of a long-standing debate in evolutionary anthropology, linguistics, and primatology. Theories bearing on this question can be said to fall into three broad categories, each of which has extensive consequences for the evolution of language (for a review see Fay and Walker, 2026). There are proponents for each of the following theories:

(1) The vocal-first theory. The main idea is that language evolved directly from primate vocal calls. Already Darwin (1871) in The Descent of Man suggested that language evolved from singing sounds used in courtship. MacNeilage (1998) is an early modern defender of the theory. Dunbar's (1996) theory of language as a form of “vocal grooming” is also of this type. Fitch (2010) and Mithen (2024) claim that vocal communication is the primary basis for human language (see also Cheney and Seyfarth, 2005). Toward this line, while it has long been argued that the vocal behaviors of primates—the closest extant relatives of humans—are mostly inflexible, the last two decades has seen a trend among primatologists away from this idea (for example, Zuberbühler, 2011; Lameira et al., 2016; Taylor et al., 2022; Ekström et al., 2024). Such evidence has led some to revisiting the usefulness of primates as model animals in debates over the evolution of vocal communication, arguing that primate vocalizations are more flexible than previously thought and may have provided a foundation for structured sound (for example, Lameira et al., 2025).

(2) The gesture-first theory. This theory argues that hominins communicated through gestures before developing the neurological and anatomical capacity for speech. One of the first proponents was Rousseau (1781). Hewes (1973) is a strong proponent, citing evidence from nonhuman primate studies and Neanderthal physiology. Corballis' (2002) book develops the theory. Tomasello (2010) also argues that human cooperative communication began in the gestural domain, with roots in pointing and pantomiming (see also Zlatev et al., 2020).

(3) The multimodal theory. This theory posits that language emerged from a complex, integrated system combining hand gestures, facial expressions, and vocalizations. McNeill (1992), Bickerton (2007) and Holler and Levinson (2019) advocate the view that language is fundamentally multimodal, with speech and gesture being part of a single, unified system. Arbib's (2012) Mirror System Hypothesis also falls into this category.

Perniss et al. (2010) point out that neither vocal-only nor gesture-only explanations for language origin can explain why the other form of communication also exists and how it arose. They argue that the neural systems controlling vocalization and gesture are tightly integrated because these systems have been interconnected from the beginning. This position is supported by developmental and neurophysiological works indicating the role of Broca's area in the organization of basic motor tasks. In the brain of an infant this area is used both for manipulation and for controlling language. At the end of the second year, the area is split into two parts where only the lower is specialized for advanced language functions. This relatively late split indicates that language has an embodied foundation. This assertion is supported by work on embodiment and multimodal meaning construction. In one recent example, (Khatin-Zadeh et al. 2025) found that participants judged metaphorical sentences significantly faster and rated them as more sensible (as metaphors) when accompanied by congruent gestures matching the metaphor schema (e.g., upward motion: prices rising). As such, the study pointed to an embodied integration of speech and gesture facilitating the construction of meaning (see also e.g., Gibbs et al., 2004; Johansson Falck and Gibbs, 2012; Casasanto and Gijssels, 2015; Ekström et al., 2022).

As background to the gesture-first and the multimodal theories, several authors have claimed that the evolution of communication went through several stages, from a signaling system via pantomime and protosign and protospeech to a language with syntax (see, for example, Dessalles, 2007; Arbib, 2012; Gärdenfors, 2020). Protosign emerges when the pantomimes are broken down to conventional units (“words”) that are recognized as such within a society. It is assumed that there has been a transition from primarily gestural to primarily vocal communication. This is the transition from protosign to protospeech. Corballis' (2002 book From Hand to Mouth describes why such a transition would constitute an evolutionarily important step.

The book highlights several advantages of spoken language compared to one based on gestures: (a) vocalization allows for communication in the dark and in environments containing visual obstacles where gestures would not be easily visible; (b) vocalization is faster to produce and less energy-costly than gesture; and (c) vocalization frees the hands and eyes for other tasks (for example, tool manufacturing and cooking). Another argument is that the auditive system is typically not as busy as the visual: one can talk while engaged in other task that require visual attention. Corballis (2014, p. 185) writes: “On this view, speech can be considered the end point of a conventionalization process in which pantomimic representations are replaced by arbitrary vocal signals.” However, this does not answer the question of how the sounds produced by the mouth acquire meaning. The purpose of this paper is to argue how sound iconicity may provide a plausible mechanism for transferring meaning to vocal sounds.

Words are indeed conventions, but a conventionalization process must have initial seeds. Burling (2005, p. 123) formulates the objection as follows: “[T]he gestural theory has one nearly fatal flaw. Its sticking point has always been the switch that would have been needed to move from a visual language to an audible one.” In our opinion, the question of whether gesture or speech came first is wrongly put (see also Kendon, 2017). Human communication has always been a combination of both, even though the balance of the elements may have shifted from mainly gestural to mainly vocal. We therefore defend the multimodal theory. Kendon (2017, p. 165) concludes that “there seems to be no reason to suppose that the development of a capacity for symbolic expression was at first confined to only one modality.” Research on gesturing has shown that gesturing and speaking are components of a single process of utterance generation (Kendon, 2004; McNeill, 2012). The question should rather be how the vocalizations and iconic gestures of our common ancestors in the primates have become a productive and mainly a symbolic form of communication (Corballis, 2014; Goldin-Meadow, 2021).

Arbib (2012, p. 231) expresses a similar position: “Neural mechanisms that evolved for the control of protosign production came to also control the vocal apparatus with increasing flexibility, yielding protospeech as, initially, an adjunct to protosign.” It should also be noted that primates can voluntarily control their lips and tongues and many of them, in particular geladas, have rich communicative systems of vocalizations based on lip-smacking (Davis and MacNeilage, 2004; Bergman, 2013). However, it is not known how the control of vocal cords evolved in the hominins.

A foundational challenge for multimodality accounts has remained the specifics of how a system of communication could shift from primarily visual to primarily vocal. Here, we argue that there are more types of non-arbitrary vocal-meaning mappings than have previously been considered. Such mappings may therefore have provided stepping stones in the transition from gestures being dominating to a more vocally centered language. As with most evolutionary theories, evidence for our suggestions is mainly indirect.

In this article, we first review several already established sources of iconicity in vocal signals—including “size symbolism” and symbolic relationships in pitch perception. We then turn to more speculative avenues, and highlight the potential roles of rhythm, loudness and amplitude, and voice quality—the potential of which have remained underexplored in theories of language evolution.¹ We conclude by summarizing identified correspondences and argue that they may provide “gestural accounts” with a gradual pathway which has long been missing from such theories. We investigate the hypothesis that the shift from hand to mouth exploited different forms of sound-meaning iconicity. To a significant extent, we support Imai and Kita's (2014) sound symbolism hypothesis although we apply it in an evolutionary setting rather than in a developmental one.² The embodied aspects of iconicity are also emphasized. We argue that the shift from gestures to spoken signs was a gradual process driven by replacing gesturally mimed patterns with iconicity based on various dimensions of vocally generated sounds. We do this by presenting several different ways that vocal sounds iconically exhibit meaning.

In contrast to a symbol, an icon is a sign that resembles the representation denoted by the sign. For example, a silhouette of a woman on the door to the ladies' room is an icon for the persons who are supposed to enter. In distinction to symbols, the choice of icons is not arbitrary, but dependent on a similarity between the sign and what it denotes. Many gestures are iconic. Iconicity, however, comes in degrees. Signs evolve by repeated use in communication. Several experiments have analyzed this process (Fay et al., 2013; Nölle and Galantucci, 2022). Similar processes can be found among non-human animals in ritualizations of gestures. A difference is that animal ritualizations occur in dyads of individuals, mainly in the interactions between infants and mothers, while in humans the conventionalization of sign is a social process. The process is driven by increased efficiency. For example, Placiński et al. (2023) performed a motion capture experiment suggesting that, over repeated social interactions, the efficiency of gestures is due to a change from whole-body pantomimes to more efficient manual gestures.

Because of its root in similarity, an icon can often be interpreted through general knowledge of the world, without previous agreement with others. There exist, however, many kinds of similarity and hence, as we shall show, there are many kinds of voice iconicity. The study of such cross-modal mappings is relevant for the question of how interactions between gestural and vocal communication evolved in ancestral primates and humans. Modern humans readily pantomime gesturally, that is, represent a sequence of events iconically with gestures, but this ability is extremely limited in other species. Humans also pantomime vocally. We hypothesize that vocal pantomimes may gradually have replaced gestural ones.

It may seem that the possibilities for iconicity in speech are very limited compared to sign. For example, Burling (2005, p. 115) writes that “the three spatial dimensions within which signs are produced invite a degree of iconicity and indexicality that is impossible in the single dimension of time to which spoken languages are confined.” It is incorrect, however, to describe speech, let alone sound in general, as one-dimensional. One contribution of our article is that we give evidence that speech has several dimensions which support various kinds of iconicity. In the next three sections, we present these forms of iconicity. Another contribution is that we make it plausible that pantomime can be transferred from the gestural to the vocal mode.

2 Sound-meaning iconicity

It has often been claimed that vocalizations by themselves lack the potential to express semantic information through their form, whereas gestures provide comparatively rich material for representation (for example, Hockett, 1978, p. 274; Tomasello, 2010; p. 228). However, evidence is now accumulating to support the idea that vocalizations can be iconic (e.g., Perlman and Cain, 2014; Perlman and Lupyan, 2018; Cwiek et al., 2021; Sidhu, 2025). Beginning with Köhler (1929) mapping of non-sense words “takete” and “baluma” to angular and curvy shapes, respectively, several additional forms of iconicity, defined as statistically significant non-arbitrary mappings between form and meaning, have been established also for non-verbal vocalizations (Ahlner and Zlatev, 2010; Perlman and Cain, 2014; Perlman and Lupyan, 2018; Pisanski et al., 2022; Ekström et al., 2022). The discussion of iconicity carries significant weight for research on the evolutionary origins of spoken language (Perlman and Cain, 2014). Once the hominin vocal system had been reshaped to allow greater phonetic variations, new possibilities for vocally expressed meanings opened up. Indirect support comes from studies showing that iconicity facilitates language learning (Imai and Kita, 2014; Nielsen and Dingemanse, 2021).

The question of whether there is a natural connection between sound and meaning has a long history. It is becoming clearer that there are several types of non-arbitrary connections. In this section, we present some well-known types of sound-meaning iconicity, following the classification of Svantesson's (2017) survey article. In this framework, onomatopoeia is perhaps the most classical example of sound-meaning iconicity. It seems that all languages have onomatopoetic words. Dingemanse (2012) defines “ideophones” as “words that depict sensory images.” Svantesson (2017, p. 3) gives as an example the Kammu word “cáac-càac,” meaning “walk.” Other examples are bird names (for example “cuckoo” and “chiffchaff” in English).

A second type is size iconicity. The most well-studied example is that “bright” vowels such as /i/ and /e/ occur in words that are used to denote small objects, while “dark” vowels such as /o/ and /u/ occur in words denoting large objects. There are two main explanations for this phenomenon (Sapir, 1929). The first is acoustic, hypothesizing that high frequencies represent “small” while low frequencies represent “large.” Essentially, because on average, smaller creatures will cause higher-frequency vibrations than larger ones, this relationship is easily inferred (Ohala, 1984). While there is an intuitive interpretation that a “larger [oral] chamber” maps onto “dark” back vowels /u/ and /O/ (the vowels in the words “pool” and “thought,” respectively, both of which involves enlarging the front cavity), modeling shows more complex rules and relationships (Fant, 1971). Specifically, a lower second formant, as in back vowels, acts as secondary cue to “bigness”—but there are in theory many configurations of vocal tract gestures that may achieve such effects (Song et al., 2026).

Finally, there is also shape iconicity. The classic example is Köhler (1929) example of the words “takete” and “maluma.” He found that readers tended to associate “takete” with pointed objects and “maluma” with rounded ones. The same effect was found by Ramachandran and Hubbard (2001) with respect to the words “kiki” and “bouba”—which has become the standard for the field. It has been shown that even two-and-a-half-year-old children are sensitive to the kiki-bouba effect (Maurer et al., 2006). Indeed, results reported by Erben Johansson et al. (2021) suggest that such sound-shape mappings may emerge and be amplified culturally through learning across generations. More recent evidence has shone further light on the origin of the mappings themselves; results reported by Lev-Ari (2026) suggests that it is the relative distinctiveness of stimuli that is targeted in sound symbolic non-arbitrary attribution, rather than the absolute characteristics of stimuli, suggesting broader processes of cognitive category formation may play an, as of yet, unrecognized role in explicating iconic relationships.

3 Prosody-meaning iconicity

In this section, we focus on different kinds of iconicity based on prosody—an area that has been little studied. In speech sciences, prosody is defined as a suite of dynamic phonetic cues, from changes in speech pitch to the stress, rhythm, and loudness of an utterance or segments of an utterance. Each prosodic aspect lends itself to iconic inference.³ Vocalizations in animals and infants are almost always expressing emotions, and do so rapidly and readily (e.g., Banse and Scherer, 1996; Pell and Kotz, 2011; Pell et al., 2026). It seems that prosody is central for emotive expressions—for example, even prairie dogs exploit difference in prosody in their communication (Slobodchikoff et al., 2009), and vocalizations by common domestic cats will exhibit variable pitch curves based on situational needs (Schötz et al., 2024).

The question of what selection mechanisms caused the evolution of the control of vocal cords in the hominins is still not answered. Indirect support for the importance of prosody comes from Nishimura et al. (2022; see also Lameira et al., 2016) who study a wide range of extant primates, combining observations of phonation with modeling. The hominins gradually lost the air sacks around the larynx that are found in other apes. The loss of these tissues enabled humans to have excellent pitch control with long and stable speech sounds generates relatively stable fundamental frequency f₀ which allows the human larynx to highlight formant changes that convey most phonetic information. The air sacks allow other primates to make louder, higher-pitched calls than humans, but they make voice breaks and lead to noisy vocal irregularities. Somewhat paradoxically, the evolutionary simplification of our laryngeal anatomy made possible the increased complexity of human spoken language (see also Lieberman, 1984; Ekström et al., 2025).

An early study concerning the role of prosody in emotional expressions is Fernald (1992) who studied how mothers and their infants in different cultures communicate. Of course, the mothers use different phonemes, different words and different nursery rhymes when talking to their infants. However, it turns out that the prosody exhibits four fundamental structures: (i) The mother can encourage the infant to do something (“can you get it?”) with rising pitch. (ii) She can reward the child for something it has done (“good girl”) with falling pitch. (iii) She can warn or forbid the child and then the phrasing is often short sound thrusts (“no, no, no”). (iv) Finally, she can comfort the child and then the melody of the speech shows a soft, billowing pattern.

These four sound patterns seem to occur in all cultures, independently of which language is used to express the messages to the infants. (A link to animal communication is that in many species warnings consist of sound thrusts). Interestingly enough, Fernald's results to some extent justify Darwin's song theory of the evolution of language. The important feature is the emotional content, rather than the semantic meaning of the words that are used. This further supports the notion that a fundamental function of language is to express feelings.

Donald (2001, p. 30) writes about infants' learning of language:

They know about the “prosodic,” or emotional, features of speech a very long time before they can understand its meanings. Later they use this knowledge as an attentional hook to help them home in on the intended meaning of what they hear. The speaker's tone of voice tells the infant where and when to look and marks the significance of the utterance. The situational context usually provides the meaning. This reduces the difficulty of tracking the content hidden in words and sentences.

In a study by Ekström et al. (2022), participants were presented with two short videos, where one involved an object rising in space and the other an object descending in space; presented simultaneously with each video in a trial was the same two-syllabic nonsense word, resynthesized with either an increasing or a declining fundamental frequency (f₀) prosodic trajectory (corresponding to an increasing or lowering rate of vocal fold vibration, and correspondingly higher or lowered pitch to listeners). Participants were tasked with choosing which of the two videos the word described. Results of the experiment indicate that prosody is iconically associated with motion, primarily driven by the selections of words with declining f₀. Further experiments supporting prosody-meaning iconicity have been performed by Suzuki and Nagai (2024), where participants were required to produce a high- or a low-pitched vowel in response to an upward or downward movement on a screen. The results show that the reaction times were faster when the pitched corresponded to the motion (high pitch/upward and low pitch/downward) than when it did not.

The results of these experiments suggest a pitch-motor correspondence between the motion and the speech process. It can be seen as support for the multi-modal approach to language evolution since not only gestures, but also vocal sounds can metaphorically represent motion.

4 Further dimensions of sound iconicity

Speech acoustic signals contain multiple features which have thus far remained relatively understudied in iconicity research—especially in the context of language evolution. These aspects may, however, in theory have been applicable in facilitating a shift from mainly gestural to mainly vocal communication. In this section, we expand on three such aspects—rhythm, loudness and amplitude, and voice quality—all of which closely mirror both physiological and emotional states It may be thought that the vocal features we discuss in this section function as indices of effort, arousal, affect, or bodily state rather than as iconic features. However, even though the vocal features may originally have been indices, they became part of an embodied iconic system when they were used in intentional communication.. We note that empirical data for these potential iconic relationships are limited and our accounts are therefore more speculative.

4.1 Rhythm

One potential source of iconic content is speech rhythm, which in speech corresponds to temporal organization of elements into patterned stress, pause and tempo. We are, however, aware of few works that have investigated ostensibly iconic properties of rhythmic speech (for example, Leong and Goswami, 2015). However, it is known that human speakers modulate speech tempo to describe, for example, speed (Shintel et al., 2014; Perlman et al., 2015). Broadly considered, however, these works have not considered the roles of speech rhythms in the context of the evolution of language.

Urgency and temporarily are readily conveyed iconically through manual hand movements, for example through the speed, rate of repetition, or “jerkiness” of gestures. We suggest that rhythmic patterning of acoustic elements or units may indeed afford iconic relationships through the emulation of internal states. Rhythms, many of which are incidental or generally involuntary, are common in nature, including rhythms inherent to breathing, or bodily movement more generally (for example, Larsson, 2014). Indeed, breathing cycles tend to broadly “organize” both calls by vocalizing animals (Schmidt et al., 2012) and speech in humans (Fuchs and Rochet-Capellan, 2021; Di Pasquasio et al., 2025). As such, rapid-fire rhythms may effectively “depict” urgency, while slower cycles may convey the opposite. Indeed, while literature pertaining to such a possibility is sparse, there is widespread evidence of rhythmic temporal structures in vocalizations across species (Ravignani, 2019), and more specifically, clusters of vocal staccato “bursts” frequently occur in contexts of distress or aggression in animal vocalizations, including, for example, bats (Hechavarría et al., 2016) and birds (e.g., Goodale and Kotagama, 2005).

The ability to couple sound sequences to hand gestures exist in non-human primates (Archakov et al., 2020), indicating a mechanism and neural infrastructure by which motion-auditory mapping may have taken place. Voluntary control of rhythmic output would, in this view, represent a relatively low-cost evolutionary exaptation, allowing vocal behavior to inherit (and later, by extension, replace) iconic functions previously served by visible pantomime. From an evolutionary point of view, thus, rhythm may have served to externalize temporal dynamics from the gestural and pantomimic modality to the vocal and auditory one.

4.2 Loudness and amplitude

A second source of potential iconicity in voice signals is the loudness of voice itself, and its potential for signaling cues effort, social dominance, and size. When gesturing, such qualities may be iconically displayed through, for example, the forcefulness of hand movements. Meanwhile, acoustic amplitude (the physical measurement corresponding to the height of the soundwave, which is perceived as loudness by human listeners) corresponds closely to vocal effort and thus offers one of the most direct forms of iconicity in vocal production behavior. Greater amplitude physiologically mirrors an increase in subglottal pressure, respiratory flow, and laryngeal tension, rendering louder vocalizations an honest index of physical exertion. Indeed, in animal communication, loudness commonly signals aggression or dominance. In this sense, manual and vocal-auditory domains provide analogous signals, where greater physiological effort is naturally mapped onto embodied meaning signaling that effort (or, in the case of for example size, potential to exert greater effort).

A key insight from recent work on loudness-frequency trade-offs (Anikin et al., 2024) offers additional potential clues. Namely, producing loud calls also involve raising f₀, favoring open vowels such as “aah,” as wide mouth opening and high subglottal pressure elevate pitch. For this reason, aggressive and intimidating calls are often surprisingly high-pitched (counter to expectations posited by the “frequency code”). “Low and loud” remains a rare, honest signal of true physical size (large body size paired with effort capacity). The mechanistic underpinnings of amplitude and loudness, thus, would theoretically provide a “gesture-first” system with a ready set of graded expressions by which distinctions in arousal, effort, social dominance, could be iconically and non-verbally conveyed. Any increases in control of the vocal organs would, in such a scenario, have allowed for further distinction of expression, including for example the height of emotion. As such, loudness iconicity could have served as a way of bootstrapping a symbolic layer onto primate vocal repertoires. In externalizing effort auditorily and conveying urgency and intensity, “loudness iconicity” may also have played a role in facilitating a shift away from a “gesture-first” scenario.

4.3 Voice quality

In phonetics, voice quality broadly refers to distinct modes of vocal fold activity in speech production. A “modal” quality refers to neutral phonation characterized by mostly regular vocal fold vibration, as heard in normal conversational speech. In contrast, for example, a breathy voice corresponds to a partially open glottal opening, facilitating greater airflow and a correspondingly airy quality. A creaky voice, finally, involves adduction of vocal fold tissue, with lower-frequency and low-regularity vocal fold vibration. As the acoustic phenomena of voice quality are tied to physical states of the vocal folds, we argue, they too may provide an elementary form of “gateway” iconic affordance for vocal behavior more broadly.

Specifically, deviations from regular (modal) voice qualities such as breathiness and creak lend themselves to non-arbitrary mapping onto emotional states. Breathiness, for example, is generally found to be the most attractive voice quality in experimental settings—that is, mapping onto settings of intimacy and trustworthiness (Anikin, 2020), while creak maps onto antagonism or authority (Yuasa, 2010; Greer and Winters, 2015; though note this is likely mediated by gender, per, for example, Anderson et al., 2014). In animal communication, vocalization patterns often follow similar schemata. For example, in domestic cat calls, inconsistent or creaky cues are associated with threat displays (such as growling), while “yowls” are characteristically strained, akin to human crying or sounds of distress (Schötz, 2020, p. 321). Meows—the type of vocalizations most used by cats to communicate with humans (for example, Bradshaw and Cameron-Beaumont, 2000; Tavernier et al., 2020)—are generally produced with a register most closely corresponding to modal voice, the natural vocal register employed by humans in everyday speech; i.e., produced in a relatively relaxed physiological state.

While relatively few works exist that investigate iconicity in voice quality, experimental studies by Akita (2021, 2025) are consistent with the interpretation that harsh and pressed voices may convey aggressive intent. Intriguingly, in Akita's (2021) study, voice quality affected perception of words more consistently across languages than did consonants and vowels in a word. “Creak”—defined by inconsistent vibration rates—was perceived as “spiky,” while breathy voices made words appear “softer.” The sum of relevant findings thus suggests that iconic relationships rooted in laryngeal physiological states (that is, physical and perceptual dimensions of vocal fold activity) may have served as a useful basis for increasingly complex vocal communication (see also Kreiman, 2024). The increased vocal flexibility would support the shift to more speech in communication.

5 Articulatory movement iconicity

The previous sections have presented iconicity in relation to aspects of speech production. There is, however, another directly embodied perspective that may be part of an explanation of how protosign evolved into protospeech. The motor theory of speech perception is a classic theory of speech perception, which originally argued that sounds are perceived in terms of how they are produced, rather than in terms of their acoustic characteristics (later refined in Liberman and Mattingly, 1985; see also Studdert-Kennedy, 2005).⁴ The link between cognitive and perceptual representations of speech, and its production (originally conceived of as speech being perceived by way of mentally simulating the speaker's vocal tract movements—gestures—during speech production), is followed up on elsewhere in the literature. Following MacNeilage (1998), Studdert-Kennedy and Goldsmith (2003) propose that the mammalian capacities for sucking, licking and chewing have been exapted as pantomimed gestures that have evolved into protosyllables. Indeed, recent evidence from ultrasound studies suggests that production-cognition links may be present from very early stages of life.

Frota et al. (2026) found that, despite not yet possessing the capacity to fluidly produce certain heard consonant sounds (Vihman, 2017; Ekström, 2022), human infants still produce minor tongue movements corresponding to their appropriate place of articulation. What's more, such connections may be wholly multimodal. In recent work performed by Imai et al. (2025), deaf and hearing adults performed accurately on word-shape iconic correspondence tasks. However, when articulatory movements were restricted, the performance of deaf participants markedly decreased, while hearing adults' performance was largely unaffected. These data align with a broader line of findings in relevant speech perception literature, which begins with the classic observations of so-called “McGurk” effects (first reported in McGurk and MacDonald, 1976); that is, the observation that the simultaneous presentation of incongruent speech audio (a person speaking “bah”) and visual feedback (a person speaking “gah”) facilitates an altogether distinct percept (“dah”). Broadly, the perception—and indeed production—of speech as speech is not a matter of acoustics alone but also includes embodies factors.

This assumption forms the basis for sound-action iconicity. Vainio and Vainio (2021) present several results linking actions and speech sounds. For example, they show that the vowel /i/ is associated with the precision grip (where an object is held between thumb and index) and /a/ is associated with the power grip (where an object is held with the whole hand). The explanation they give is the same as above, that is, /i/ corresponds to a small vocal cavity, while /a/ corresponds to large one. Such mappings suggest shared sensorimotor mechanisms between vocal articulation and manual action, potentially rooted in evolutionary exaptation of motor representations (Vainio et al., 2024a,b). Another result is that these effects extend to spatial concepts, with high pitch linked to “open” or “up” and low pitch to “closed” or “down,” further bridging gestural and vocal iconicity.

6 Conclusion

There are yet many unanswered questions about how the uniquely complex human vocalization abilities evolved. So-called “gesture-first” theories have long been critiqued on the basis that no clear trajectory would allow the transfer of meaningful communication “from hand to mouth,” that is, the transition from protosign to protospeech. Both gestural and vocal iconicity are likely influential factors in the emergence of human communication systems. An example of the complementary influence of words and gestures is provided by Khatin-Zadeh et al. (2025). They showed that when gestures congruent with a verbal metaphor were presented, the acceptability increased and the reaction times decreased, compared to when an incongruent gesture was combined with the metaphor. Their results provide further support for a multimodal grounding of language.

The aim of this article has been to argue that sound iconicity has been an underestimated factor in this transition. One contribution is that we have highlighted several possible iconic relationships in vocal signals. We have presented a set of embodied vocal affordances which may have facilitated the transition from protosign to protospeech. We have also argued that pantomime need not be gestural but can also be based on vocal similarity. We have illustrated this with prosody iconicity. Our analysis supports a multimodal theory of language evolution rather than a purely gestural theory. A speculation is that controlling the pitch of vocalizations was one of the first adaptations of the human voice. In our opinion, prosody-meaning iconicity particularly deserves further study. However, once a conventional system of signs is established, the role of iconicitydiminishes, and selective pressures for more efficient signs may drive the evolution of communication toward a more symbolic system.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Hassan Banaruee, University of Education Weingarten, Germany

Reviewed by: Bradley Walker, University of Western Australia, Australia

Chuck Bradley, Independent Researcher, San Francisco, CA, United States

Footnotes

We focus on the physical aspects of the vocal sound. Another approach to sound iconicity is to make cross-linguistic comparisons to find out whether there exist common mappings between different sound and semantic aspects of words. See Erben Johansson (2026) for a review of this area.

Imai and Kita's (2014) hypothesis is that “(i) pre-verbal infants are sensitive to sound symbolism, due to a biologically endowed ability to map and integrate multi-modal input, (ii) sound symbolism helps infants gain referential insight for speech sounds, (iii) sound symbolism helps infants and toddlers associate speech sounds with their referents to establish a lexical representation and (iv) sound symbolism helps toddlers learn words by allowing them to focus on referents embedded in a complex scene”.

Various such aspects are, as pertaining to our theorizing in this paper, underexplored in the relevant literature; these are explored in the next section. Here, for the purpose of this section, prosody is taken to be synonymous with pitch and variations thereof.

The theory itself has been subject to extensive dispute within broader phonetic sciences (e.g., Blumstein and Stevens, 1979; Galantucci et al., 2006)—a debate with which we do not engage with in this article. Our acknowledgment of it here is mainly historical, as an early influential truly multimodal theory linking cognitive and productive aspects of speech behavior.