Constraints on communicating the order of events in stories through pantomime

Abstract

Pantomime is a means of bodily visual communication that is based on iconic gestures that are not fully conventional. It has become a key element in many models of language evolution and a strong candidate for the original human-specific communicative system (Zlatev et al. 2020). Although pantomime affords successful communication in many contexts, it has some semiotic limitations. In this study, we looked at one of them, connected with communicating the order of events in stories. We assumed that pantomime is well-suited for communicating simple stories, where events are arranged in chronological order, and less so for communicating complex stories, where events are arranged in a non-chronological order. To test this assumption, we designed a semiotic game in which participants took turns as directors and matchers. The task of the directors was to mime a story in one of two conditions: chronological or non-chronological; the task of the matchers was to interpret what was mimed. The results showed that the chronological condition was easier for the participants. In the non-chronological condition, we observed that initially, poor communicative success improved as the participants started to use various markers of event order. The results of our study provide insight into the early stages of conventionalisation in bodily visual communication, a potential first step towards protolanguage.

1. Introduction

In recent years, pantomime has often been proposed as a likely candidate for ‘the original human-specific communicative system’ (Zlatev et al. 2020): a stage in the evolution of human cognition and communication that depended on iconic (i.e., resemblance-based) gestures expressed with hands, arms, and the whole body, as well as other semiotic resources like vocalisations (cf. Żywiczyński et al. 2018). There are three lines of thought arguing for pantomime as a precursor to language that has become particularly influential. One is Donald’s mimesis theory (Donald 1991, 2001; later, e.g., Zlatev 2014; Gärdenfors 2017; Zlatev et al. 2020). A second is Arbib’s Mirror System Hypothesis (Arbib 2005, 2012, 2016; Arbib et al. 2018), and a third is Tomasello’s (2008) proposal that language was bootstrapped by two types of ‘natural human gesture’: pointing and pantomiming. There are important differences between these positions, for example regarding the question of whether pantomime was accompanied by vocalisation (Zlatev et al. 2017; Zlatev et al. 2020) or not (Arbib 2012, 2013). Still, to some degree, it is possible to identify recurrent elements in their understanding of pantomime, whereby it is dominated by non-vocal, mimetic, and not fully conventionalised means of communication (cf. Arbib’s notion of ad hoc pantomime, 2012), which is executed by coordinated movements of the body.¹ Semiotically, individual pantomimes are iconic representations of actions or events which may be displaced from the here and now (cf. Gärdenfors’s notion of detachment 2003; Żywiczyński et al. 2018). Pantomime is topically unconstrained; its capacity to cover a range of topics can be exemplified by the thematic scope of the so-called pantomimic fossils. For instance, in first-contact communication, pantomime has been used to signal intentions and plans for the future, and to transfer information about topography or trade (Żywiczyński et al. 2021: 4).

The communicative effectiveness of pantomime has been explored in studies based on communication games (e.g., Zlatev et al. 2017), where participants use pantomime to show actions and changes of state, predominantly by means of ‘acting as if’, or the ‘enacting mode’ (Müller 2014; Fig. 1).

In such studies, participants used pantomime to communicate events involving an Agent and a Patient (Zlatev et al. 2017; Gärdenfors et al. 2018; Żywiczyński et al. 2021). For example, in a study by Zlatev and colleagues (2017), pantomime was used to communicate four transitive actions: ‘kissing’, ‘pushing’, ‘slapping’, and ‘waving’. In similar studies based on communication games, typically, the repertoire of events communicated is contextually constrained by a closed meaning matrix (Galantucci and Garrod 2011; Nölle and Galantucci 2022). For instance, Żywiczyński et al. (2021) used a set of images out of which participants had to choose the one that had been mimed by an actor. As these studies show, pantomime can be successfully used to communicate not only single events but also sequences of events that are causally connected (Fig. 2; see also Ferretti 2021; Gärdenfors 2021).

Given that pantomime is based on iconic gestures, as stated at the onset, its comprehension does not rely on semiotic conventions, but on the resemblance between an expression and what it represents (Fig. 3). This resemblance may reside in shared qualities (such as shape) as well as shared relations (such as positioning in space), corresponding to imagic (or imagistic) and diagrammatic iconicity, respectively (Jakobson 1965; Peirce 1974). A classic example of diagrammatic iconicity that has to do with linguistic temporality is that of ‘veni, vidi, vici’, where the order of the verbs reflects the natural order of the events. Thus, we interpret the famous sentence to mean: first, he came; second, he saw; third, he conquered (Simone 1995; for a discussion of diagrammatic iconicity in connection with distance, see Devylder 2018). Similar examples can be found in bodily visual communication, for instance in signed languages (Sallandre and Cuxac 2001; Pietrandrea and Russo 2007). Since pantomime is strongly based on iconicity, the order in which a sequence of events is shown can be expected to match the so-called ordo naturalis, so that if a person mimes a woman eating a sandwich, then a monkey stealing the sandwich, then the woman screaming at the monkey in anger (cf. Fig. 2), we would not assume that screaming happened before eating.

Pantomime is also prototypically a ‘performance’, a form of face-to-face interaction. As such, it unfolds in real-time, one movement after another. It is thus often assumed (Kuhn and Schmid 2013) that as long as pantomime is not, for instance, video-recorded and edited, it does not easily allow for doing away with the constraints of (diagrammatic) iconicity (Ryan 2012; Żywiczyński et al. 2018: 311).

The temporally iconic character of bodily visual communication is taken for granted to the extent that it is scarcely discussed in gesture studies. Instead, many studies concentrate on constituent order (e.g., Goldin-Meadow et al. 2008) or on gesture–speech correspondence for time metaphors and to the use of space for time references (e.g., Burns et al. 2019). We know from the former that, depending on the type of event, people produce ‘silent gestures’ corresponding to different word orders in spoken language: Subject–Verb–Object (SVO) versus. Subject–Object–Verb (SOV) (Hall et al. 2013; Schouwstra et al. 2019). But these ordering patterns are applicable to elements within events, not to events as such. As far as the complex ordering patterns of whole events are concerned, we know that in bodily visual communication, it is possible to refer to different temporal planes, at least ‘the past’, ‘the present’, and ‘the future’, by means of meaning-making mechanisms other than iconicity itself. For instance, deixis, which is used, ‘to reference locations, items, or people in the world’ (Novack and Goldin-Meadow 2017: 1), offers some possibilities for referencing past or future events, regardless of the sequence of events shown. Deictic gestures can call for knowledge of cultural conventions connected with mapping time onto space, such as ordering events on a lateral (Rodriguez 2019: 7) or a sagittal axis (Núñez and Sweetser 2006: 437; Casasanto and Jasmin 2012; Cooperrider, Núñez and Sweetser 2014). In principle, it should be possible to incorporate the same or similar strategies for ordering and reordering events in pantomime.

There is at least one communicative context in which the ability to represent events in different orders is crucial: storytelling. At least since antiquity, the order of events has been discussed as an essential aspect of storytelling, with much emphasis put on its manipulations (see Abbott 2011). For instance, Horace in his Ars Poetica, advised authors to begin stories ‘in the middle of things’ (in media res) rather than start ‘from the egg’ (ab ovo) (Richardson 2002: 1). Much later, narratologists such as Genette (2002) claimed that the most common way to tell a story is starting in the middle and then introducing some elucidatory analepses, that is explanatory flashbacks. The claim about anachrony of stories can also be identified, in different wording, in the works of other, both classical and cognitive, narratologists (see ‘dechronolisation’ in Bremond 1966; Greimas 1966; ‘fictional time’ in Ronen 1994; ‘order’ in Friedemann 1910; Schmid 2010; see also Herman 2009; Sternberg 2001). Importantly, distortions of chronological order of events can be observed both in written or ‘regularised’ oral stories, such as myths and folktales, whose contents and structures are to an extent fixed, as well as in casual, conversational storytelling, which is characterised by multiple ‘false starts’ and ‘corrections’ (Norrick 2007) that can cause chronological disruptions.

Storytelling is an essential component in all human cultures (e.g., Boyd 2009; Gottschall 2012), a human-unique trait (e.g., Niles 1999), and could have played a key role in the evolution of language, for instance in the transition from mimetic to mythic culture, as per the theory of Donald (1991). According to Donald, stories (and, in particular, myths) appeared alongside or after language, taking on the form of verbal narratives (Donald 2001; Collins 2013). But that does not exclude the emergence of ‘mimetic preverbal narratives’ (Boyd 2017: 2). A narrative can be roughly understood as a representation of a story, where patterns of events and/or changes of state are reported (Ryan 2007). Mimetic preverbal narratives can be envisaged as products of sharing stories before fully-fledged language was available, by means of body movements that were imitative and intentional. The events in a story could be represented by either re-creating what happened exactly as it happened or by re-enacting what happened selectively (cf. Żywiczyński et al. 2018: 312).

There are different lines of argumentation for preverbal narratives. First, stories can be successfully shared non-verbally, with the use of pantomime and/or other semiotic resources (McBride 2014; Sonesson 1997; Sibierska 2017). Further, at least selected aspects of the cognitive infrastructure for stories are independent of language: for instance, as Ferretti and colleagues argue, global coherence, inherent in what they call ‘narrative thought’, pre-existed speech (Ferretti et al. 2017, 2021). This argument is supported with the research on story comprehension in speech-impaired populations (e.g., Autism Spectrum Disorder in Ferretti et al. 2017; Manfredi et al. 2020). At the same time, stories seem to belong to the same socio-cognitive niche as language, as a booster of cooperation and pro-social attitudes (Boyd 2017; Smith et al. 2017). The connections between stories and language are still very much under discussion. And so are their evolutionary timelines. In theory, narratives could have emerged even prior to the evolution of (proto)language. The question is about the complexity of such stories (Boyd 2017), and in particular their chronological complexity. Could stories expressed in pantomime, for example, start ‘in the middle of things’, reporting events in non-chronological order? Or does this imply the need to go beyond pantomime and introduce conventional, that is explicitly or implicitly agreed upon, signs: symbols, and thus transcending pantomime beginning to establish a protolanguage (Zlatev et al. accepted)?

2. Study

2.1. Design and Hypotheses

To address the above questions, we looked more closely at the constraints of pantomime on communicating events in a sequence in the context of storytelling. Based on the considerations outlined above, we assumed that pantomime would be suitable for communicating simple stories in which events are arranged chronologically. However, it would be less successful when people need to represent events in a non-chronological order, which is characteristic of more complex stories (Boyd 2017; Zlatev et al. accepted). To test this hypothesis, we designed a study based on ‘semiotic games’ (e.g., Krauss and Winheimer 1966; Fay et al. 2010; cf. Fay et al. 2013, 2014; Nölle et al. 2018). Specifically, we adapted a referential communication task paradigm, where the participants are asked to communicate with each other in a specific medium (e.g., drawings, gestures, vocalisations), but without the use of spoken or written language. In a number of interactions, often referred to as ‘games’ and/or ‘rounds’, the participants create novel conventions in order to communicate successfully (e.g., Galantucci and Garrod 2011; Nölle and Galantucci 2022). We invited participants to play a charades-like game, in which they had to use pantomime to communicate short stories to one another in two conditions: chronological and non-chronological. We expected that communicative success would be greater in the chronological condition than in the non-chronological condition (hypothesis 1). Further, we predicted that the participants would develop strategies to communicate a disruption of chronology, which would lead to increased success in communication across successive games in the non-chronological condition (hypothesis 2). However, such strategies would involve introducing conventionalised signs, and thus transcending the borders of pantomime into protolanguage.

2.2. Stimulus and Procedure

Fifty-two participants (female = 35, male = 17), whose mean age was 25.48 (SD = 6), were recruited from students and employees of Nicolaus Copernicus University in Toruń, Poland. They were all native users of the Polish language with no specialist training in motor-visual media, such as dance or mime, and with no knowledge of signed languages. They took part in the experiment for course credit or remuneration in kind. The participants signed informed consent and GDPR statements and agreed to be video-recorded for the purposes of the study. They worked in 26 pre-formed pairs, in which they arranged themselves according to their preference. The experiment was conducted with the use of E-Prime 3.0 (Psychology Software Tools 2016) stimulus presentation software, paired with two DELL screens, with a resolution of 1920 × 1080, placed on opposite sides of the experiment space. An experimental session consisted of four games, 12 rounds each. In each round, the two participants were positioned on opposite sides of the experiment space (Fig. 4).

One of the participants took on the role of the director, while the other one was the matcher. First, the director was shown a verbal representation of a story in Polish on screen 1 (Fig. 4) taken from the pool of 24 stories shown in Table 1. Each story consisted of three events. Half of the stories had a woman as the main character while the other half had a man as the main character. A total of 12 stories contained transitive actions, such as eating, lifting, or reading, in the first event. A total of 12 stories contained intransitive actions, such as jogging, swimming, or walking, in the first event. The contents of the stories differed: 9 stories had only 1 character while the remaining 15 stories had 2. Six stories included an animal, such as a bear, a monkey, or a shark, and 16 stories included a prop, such as a bottle, a razor, or a stone. Five stories included explicitly expressed emotions: for example, ‘shouted in anger’. Overall, the pool of stories could be considered a complex meaning space with multiple feature vectors (cf. Kirby 2007).

In the chronological condition, the story was arranged into two or three separate sentences, with the third event expressed at the end, which corresponded to the ‘natural’ order of events (ordo naturalis). In the non-chronological condition, the story started with the third event, and then a subordinate clause including the Polish equivalent of the word ‘because’ (‘ponieważ’) introduced events one and two. The two conditions are exemplified in Table 2.

In each game, there were six stories in the chronological condition and another six in the non-chronological condition. The stories and their order were randomised for each game. The same story could not appear twice in the same game, but it could re-appear in successive games. After reading the story, the director was to communicate it to the matcher in a pre-specified space, where they were visible to the matcher on the opposite side of the room (Fig. 4), and, at the same time, video-recorded with a Panasonic HC-V700 camera mounted on a tripod. The director was instructed to use whole-body movement and not to use speech or non-verbal vocalisations; still, some participants spontaneously deviated from the instructions and produced various sounds.

After watching the director, the matcher was presented with four comic strips displayed on screen 2 (Fig. 4). The strips were placed from top to bottom and read from left to right. One of the strips matched the story from the director’s input in the same order. One of the other three corresponded to the story that was being communicated, but in the opposite order, and two corresponded to one of the other stories in two different orders (Fig. 5), randomly selected for each round by the software. The order in which the comic strips were displayed to the matcher was also randomised.

The task of the matcher was to choose the alternative that corresponded to the story that was being communicated by the director through pantomime in the right order: chronological or non-chronological, depending on the condition. After the matcher made their choice, both participants saw a smiling face for a correct match or a sad face for an incorrect match. The visual feedback was supported with verbal feedback from the experimenter: ‘correct answer’ or ‘incorrect answer’. Then, the participants switched roles, moving on to the next round. At the end of each game, they also got verbal feedback from the experimenter on their total score for this game. Communicative success was operationalised as the accuracy of the matchers’ responses. We recorded 1248 responses in total: 624 in the chronological condition and 624 in the non-chronological condition, which translates into 156 responses in each of the games in each of the conditions.

2.3. Coding

As stated at the end of Section 1, we hypothesised that in an interactive setting, the participants would start to communicate a disruption in chronology, for instance by inventing markers of event order (see Zlatev et. al. accepted). Based on a viewing of a random selection of ten video recordings, we developed a list of such markers. For example, the director could simply enact the events in a non-chronological order, where ‘enacting’ means mapping the director’s body onto the body of a character from the story and acting as if they were the character (Müller 2014; Zlatev et al. 2020). The director could also use a metaphorical gesture or a series of gestures to refer to the past in the space behind them; to enumerate the conditions or to enumerate the events; or to attract the matcher’s attention (Fig. 6a–c).

We employed two expert coders, independent of the main researchers involved in the study, to code the presence or absence of markers of event order in all 1248 video recordings. First, the coders worked independently and, then, compared the results of their work. There was a satisfactory agreement between the coders, kappa = 0.77 (95% CI, 0.73 to 0.81). The final version of the coding was created on the basis of consensus. There was no unsettled disagreement between the coders.

3. Results

Table 3 shows accuracy counts and proportions as well as counts and proportions of marker use by the participants. Overall, the accuracy in the chronological condition was very high. Accuracy in the non-chronological condition, on the contrary, was very low in the first game but improved gradually across the subsequent three games. The table also shows a similar pattern in the use of event order markers by the participants. In the chronological condition, the frequencies of event order marker use were comparatively low across the four games. The corresponding frequencies in the non-chronological condition were higher already at the first game and continued to increase after that. Together, these patterns suggest a positive association between the use of the markers of event order and the success with which the directors were able to communicate the content of the stories to the matchers. That is, the correct identification of stories that were presented in a non-chronological order improved if the event order was marked by the director, but for the stories presented in chronological order, the markers did not have any effect.

We analysed the data with the use of R Statistical Software (version 4.1.1; RStudio Team 2019) and the lme4 package (Bates et al. 2015). We ran a series of mixed effects logistic regression models to test the effects of ‘condition’, ‘game’, and ‘marker’ on response accuracy. Pairs of participants were included as a random effect. A series of nested models of increasing complexity was created. The simplest of these included only an intercept. After that, the predictors were added one by one to subsequent models, first as main effects, and then as two-way interactions. The most complex model that was significantly better than its predecessor (likelihood ratio test, chi-squared = 34.06, df = 1, P = .000) contained the three main effects and two two-way interactions between game and marker and between condition and marker.

Figure 7 shows the predicted proportions with 95% confidence intervals from this model. The figure adds to the observations that were made on the basis of the values in Table 3. Overall, the proportions of correct answers to chronological stories were much higher (near ceiling) than those to non-chronological stories. However, when the order of the events was marked by the directors in non-chronological stories, performance was better already during the first game and continued to improve over the next three games.

Given the significance of the two-way interactions, we compared the effects of order marker within each game in the chronological condition and in the non-chronological condition. The comparisons were done using the multcomp R-package (Hothorn, Bretz and Wesfall 2008). The P-values reported below have been adjusted for multiple comparisons. Within the chronological condition, the effect of marker was significant for game 1 (B = 1.76, SE = 0.55, z = 3.21, P =.010), but not for game 2 (B = 0.44, SE = 0.51, z = 0.87, P =.928), 3 (B = 0.01, SE = 0.59, z = 0.02, P = 1.000) or 4 (B = –0.47, SE = 0.62, z = –0.77, P =.957). Within the non-chronological condition, the difference was significant for games 1 (B = –1.77, SE = 0.47, z = –3.74, P = .001), 2 (B = –3.10, SE = 0.48, z = –6.49, P = .000), 3 (B = –3.53, SE = 0.47, z = –7.54, P = .000), and 4 (B = –4.01, SE = 0.51, z = –7.86, P = .000).

4. Discussion

The results supported both hypotheses. First, pantomimes of chronologically ordered events had a much higher communicative success (near ceiling) than pantomimes of non-chronologically ordered events. This implies that pantomime is an effective means of communicating simple stories and is less suitable for complex ones (see Zlatev et al. accepted).

As for the second hypothesis, the analysis of pantomimes in the non-chronological condition showed a significant increase in the accuracy of the matchers’ responses in the course of the experiment session. The communicative success of pantomimes in the non-chronological condition that contained markers of event order was significantly higher in the first game than that of the pantomimes in the non-chronological condition that did not contain them. When markers of event order were used, accuracy increased significantly across the four games. In contrast, the accuracy of interpreting pantomimes in the non-chronological condition that did not contain such markers stayed low throughout the four games.

We also observed an interesting pattern in the chronological condition, depending on whether they contained markers of event order or not: the communicative success of pantomimes without markers remained at a ceiling level across the games, while that of pantomimes with markers in the first game was significantly lower than the success of the corresponding pantomimes that did not make use of markers. When event order markers were present, the result improved in the second game and increased significantly in the third game, reaching the ceiling level. This suggests that when we add information, in the form of a marker of event order, about the chronological order of events in the chronological condition, this may lead to confusing the chronological and non-chronological conditions and creating noise (see ‘semantic noise’ in Jandt 2017: 82–83; cf. ‘internal noise’ in Hockett 1952; Shannon and Weaver 1949).

To sum up, the optimal strategies for successful communication in the two conditions were in opposition to one another: in the chronological condition not to use markers, and in the non-chronological condition to use markers of event order. The way in which markers of event order were used by the participants ties in with findings from research on the emergence of conventionalised signs in bodily visual communication. Both experimental studies on improvised gestural communication (e.g., Motamedi et al. 2019; Motamedi et al. 2021), as well as research on emerging signed languages and homesign (e.g., Goldin-Meadow 2003; Senghas et al. 2004; Goldin-Meadow et al. 2014), suggest that such symbolic, that is conventionalised, signs emerge gradually, in the course of horizontal and vertical transmission. In our study, we could observe an introduction of a symbolic sign at a very early stage, when a specific use of movement for communicative purposes stabilised in at least one of the participants and was successfully understood by their partner (cf. the notion of ‘usualisation’ in Schmid 2015). This can be interpreted as a first step in successful transmission that leads to the emergence of language-like structures (Motamedi et al. 2019) or in fact a form of protolanguage (Zlatev et al. accepted). At the same time, it should be noted that most of the conventionalized signs that were used as markers of event order, such as those based on enumeration, were not invented ex novo, but ‘imported’ from the shared semiotic systems of the present-day participants. This is, of course, a general methodological drawback in experimental semiotic studies, and implies the need to be careful with strong interpretations.

With this reservation in mind, the results of our study can be interpreted in the context of the early stages of compositionality in a communicative system. Compositionality allows for arranging simple units into more complex ones, whereby the meaning of the whole depends on the meaning of its parts. In signed languages, compositionality has to do with combining the parameters of movements, such as the positioning of a body part, as well as combining movements as units (Pleyer et al. 2022). Silent gestures and signs of signed languages at an early stage tend to be more holistic and become more and more compositional in later generations of participants (Kirby et al. 2022) or cohorts of signers (Senghas et al. 2004; Clay et al. 2014).² Pantomime is not compositional in the above sense, but more holistic. Its units, if there are any, are for the most part not discernible: many articulators are used at the same time, in a stream-like fashion, so that it is difficult to say where one movement terminates and another starts (Żywiczyński et al. 2018). Thus, the meaning-making processes in pantomime depend more on producing and understanding the whole pattern rather than its parts.

In our study, the way of enacting events themselves did not change across conditions; it retained the typical transitive structure with an agent, a patient, and an action, in both SVO and SOV ordering patterns (Warglien et al. 2012; Gärdenfors 2014), represented mostly in the stream-of-movement fashion that is characteristic of pantomime (Żywiczyński et al. 2018). Some pantomimes, however, were supplemented with event order markers, which had two characteristics: (1) they were used before enacting an event and/or in between events; (2) the same markers were also reused for different events and/or repeated. Thus, they functioned similarly to the silent gestures used for transfer of temporal information, which is distinct from the rest of the utterance and is usually presented at its beginning (Schouwstra 2017). The same applies to the so-called dedicated gestures in signed languages: the movements of different body parts that have a grammatical function: for example, raising brows to turn a statement into a question (Zeshan 2004). Evidence from emerging signed languages (Al-Sayyid Bedouin Sign Language in Sandler 2012; Israeli Sign Language in Meir and Sandler 2008) shows that such dedicated gestures are attached to existing signs by new cohorts of signers to add on meaning. At first, they appear spontaneously, do not have a clearly determined structure and are not strictly synchronised with the production of the sign they are ‘added to’. In later cohorts of signers, dedicated gestures tend to become more and more systematic: the position of the articulators becomes fixed and the production of the gesture becomes more closely aligned with the timing of the accompanying signs (Sandler 2012).

We know from previous experimental semiotics studies that there are certain contexts that are particularly preferential for systematicity. As Nölle et al. (2018) note, displacement is one such context: if a referent in a meaning space is displaced, a pressure on working memory appears. Increasing systematicity is a way of relieving it. This line of reasoning potentially applies to representing events in an unnatural, non-chronological order, which is more demanding for memory than following ordo naturalis.³ As the demand is increased, markers of event order might be more likely to occur and to become more systematic over time.

As pointed out above, many of the markers of event order that were observed in our study relied on shared cultural conventions to a great extent: enumerating events with the fingers, organising the miming spaces into frames of comic strips or a timeline, and using a metaphorical backwards gesture to refer to the past. Interestingly, the latter can be observed in signed languages, in which past and future are referred to alongside backward and forward movements (e.g., Jacobowitz and Stokoe 1988 for American Sign Language; Łozińska 2012 for Polish Sign Language). For instance, the sign ‘now’ in American Sign Language and Polish Sign Language make use of the space directly in front of the signer (Fig. 8a). In turn, the sign ‘tomorrow’ in American Sign Language and Polish Sign Language involves a forward movement of the hands, whereas the sign ‘yesterday’ involves a backward movement (Fig. 8b). All the participants in the sample used similar signs, which implies they all share some local conventions for representing time in space. This observation conforms with the assumption about cultural specificity in pantomime (e.g., Zlatev and Andrén 2009; Żywiczyński et al. 2021), and this applies even more to the proto-linguistic markers that emerged in the non-chronological condition due to conventionalisation.

In sum, the results of our study suggest both that pantomime as a communicative system has its limitations with respect to the representation of non-chronological order of events and at the same time that these limitations can be overcome through the introduction of compositional, proto-linguistic elements, under conditions that admittedly need to be specified more closely in future research. This lends further support for the proposal that pantomime was the original communicative system in human evolution based on sign (as opposed to only signal) use (see Zlatev et al. 2020 for a discussion of the distinction between signs and signals). These conclusions tie in with the research on the role of pantomime in the emergence of early signs in new signed languages, as stressed in particular in the context of São Tomé and Príncipe Sign Language (Mineiro et al. 2017, 2021). At the same time, this study opens up the discussion about one possible factor that could have driven the evolution of pantomime into a more language-like system: the need to convey more chronologically complex stories. Again, however, we leave it for further research to specify the ecological conditions where the need for more complex narratives would have served as an adaptive pressure.

Further, we need to be careful in drawing comparisons between the findings of our experimental study and natural communication for at least two reasons. First, the study incorporated a closed ‘meaning space’. The repertoire of events to be communicated was established by the experimenters. The directors were instructed to enact standardised stories with three events and no more than two characters (see Section 2.2). They interpreted the stories in different ways; for instance, sometimes, they omitted one or two of the three events, but, still, the information passed to the matcher was to a large extent controlled. Following a forced-choice paradigm, the matchers were instructed to give an answer with reference to the four comic strips, which further determined what information the directors had to include in their pantomimes in order for them to be correctly interpreted in two respects: (1) the story had to be distinguished from a different, randomly selected story (see Section 2.2); (2) the sequence of events had to be marked as chronological or non-chronological. Although the closed meaning space provided the communicative context crucial for pantomime (Żywiczyński et al. 2021: 7), it could affect the tendency to mime events in a particular way and use markers of event order, as we know from other research that such a referent environment does not necessarily favour the emergence of systematic strategies (Nölle et al. 2018). For an open referent environment, the tendency to use markers of event order could be even more robust. In general, a closed-meaning space casts doubt on the ecological validity of the design, as it imposes what is communicated and how. This is a problem not uncommon in similar experimental studies (Żywiczyński et al. 2021: 8). We should thus be careful with formulating conclusions about effective communication of an ‘unlimited repertoire of events’ through pantomime based on this and similar studies.

Second, the design determined the way in which the directors mimed the stories: in the non-chronological condition, the third event was expected to be shown at the beginning. In a natural interaction, this is not the only disruption of chronology that occurs in the process of sharing stories: for instance, the beginning is sometimes recounted a couple of times; the end is not recounted at all; or events are reported in a ‘mosaic-like’ fashion, where some are omitted and then supplemented not by the main storyteller, but by other interlocutors (Norrick 2004, 2007, 2018). Notably, even though the design put a constraint on the directors to show the third event first, reordering events also appeared spontaneously in ways that we did not anticipate. For example, one participant mimed the story with ID ‘bear’ (see Table 1) in the following way: first, he showed there was a bear behind a door (event 2); then, he mimed opening the door (event 1) and running away (event 3). Such a sequence did not stem from the directors’ input, nor from the concern for making the matcher’s task easier. This suggests that pantomime may to some extent express stories in a non-chronological way even if there are communicative constraints in this respect. Thus, the second hypothesis is not as unambiguously supported as suggested above (see Zlatev et al. accepted)

The extent of this tendency could be explored in a different experimental setting, possibly based on the iterated learning paradigm (e.g., Kirby et al. 2014, 2015; Motamedi et al. 2019), where a story would be passed on from one generation of participants to another, with no constraints on choosing strategies and the focus put, for instance, on its tellability, that is how salient a story is when compared to other stories. Salience can be achieved by a ‘break in expectations’, which can concern the way in which events are reported (Norrick 2004). In an experimental setting based on such a paradigm, we could see whether disruptions of chronology would occur spontaneously, that is with no direct pressure to order events in a story in a specific way; whether such disruptions would lead to a higher tellability; and whether they would prompt the participants to start to develop conventionalised (symbolic) signs for marking the order of events.

5. Conclusion

Pantomime has been argued to be a strong candidate for being the first human-unique communicative system, allowing our ancestors to express an open-ended number of messages (Arbib 2012; Zlatev et al. 2020). It even has the potential for communicating simple stories (Boyd 2017). However, it also has its semiotic limitations (e.g., Żywiczyński et al. 2021), and these need to be further explored. In this study, we focused on constraints concerning communicating the order of events in stories consisting of three events. Given the temporal diagrammatic iconicity of pantomime (of the ‘veni, vidi, vici’ type), we assume that pantomime will be well-suited for conveying simple stories, where events are arranged in a chronological sequence, but less so for expressing events in a non-chronological order, characteristic of more complex stories. This assumption, which was not tested in an experimental setting before, was corroborated in our study. Our findings showed that not only was the chronological condition easier for the participants but that communicative success was at a ceiling level. In the non-chronological condition, in turn, we could observe a learning effect, where communicative success increased with each game, and this increase was connected with the introduction of markers of event order. The results of our study provide an insight into the early stages of conventionalisation in bodily visual communication, a potential first step towards protolanguage, propelled by the need to convey more chronologically complex stories.

Acknowledgements

The authors would also like to thank the Lund University Humanities Lab for their support in conducting this research.

Funding

This research was supported by the Polish National Science Centre (NCN) under the agreement UMO-2017/27/B/HS2/00642. Marta Sibierska was supported by the Polish National Agency of Academic Exchange (NAWA) under the agreement PPN/BEK/2020/1/00319/U/00001.

Data Availability

The data underlying this article are available in the Open Scientific Framework (OSF) at https://osf.io/yg5a8/?view_only=67a48a1c60d84ca689b4c0bbb5ae974b.

References

Footnotes