1

The Meanings of Chimpanzee Gestures 1

2

Catherine Hobaiter and Richard W. Byrne 3

4

5

6

7

8

9

10

11

12

13

14

15

16

School of Psychology and Neuroscience, University of St Andrews, St Andrews, 17

KY16 9JP, Scotland 18

19

Correspondence: R W Byrne, School of Psychology and Neuroscience, University 20

of St Andrews, St Andrews, Scotland. 21

E-mail: rwb@st-andrews.ac.uk 22

23 24

2

HIGHLIGHTS 25

• 66 gestures are used intentionally to communicate 19 meanings by wild 26

chimpanzees 27

• We analyzed >4500 cases to extract true (non-play) meanings for 36 gestures 28

• Gestures have the same meaning(s) across individual signalers 29

• Flexible use of several gestures for same goal is higher during social 30

negotiation 31

32

3

Summary 33

Chimpanzees’ use of gesture was described in the first detailed field study [1,2], and 34

natural use of specific gestures has been analyzed [3-5]. But it was systematic work 35

with captive groups that revealed compelling evidence that chimpanzees use gestures 36

to communicate in a flexible, goal orientated, and intentional fashion [6-8, replicated 37

across all great ape species in captivity 9-17 and wild chimpanzees 18,19]. All these 38

aspects overlap with human language, but are apparently missing in most animal 39

communication systems: including great ape vocalization, where extensive study has 40

produced meagre evidence for intentional use [20, but see 21,22]. Findings about 41

great ape gestures spurred interest in a potential common ancestral origin with 42

components of human language [23-25]. Of particular interest, given the relevance to 43

language origins, is the question of what do chimpanzees intend their gestures to 44

mean; surprisingly, the matter of what the intentional signals are used to achieve has 45

been largely neglected. Here we present the first systematic study of meaning in 46

chimpanzee gestural communication. Individual gestures have specific meanings, 47

independently of signaler identity, and we provide a partial ‘lexicon’; flexibility is 48

predominantly in the use of multiple gestures for a specific meaning. We distinguish a 49

range of meanings: from simple requests associated with just a few gestures, to 50

broader social negotiation associated with a wider range of gesture types. Access to a 51

range of alternatives may increase communicative subtlety during important social 52

negotiations. 53

54

4

Results 55

In animal communication, signal meanings have generally been identified with the 56

information exchanged between individuals [26,27]: here, only the characteristic 57

effect of a signal on recipients is assessed. For example, monkey alarm calls function 58

as if they referred to specific predators: recipients act appropriately upon hearing the 59

calls [28-30]. Whether callers intend to influence a specific audience is unknown, and 60

suspected not to be the case [31]. In human communication, however, meaning has 61

been treated quite differently because signals - linguistic utterances - are produced 62

intentionally [32]. Indeed, the signaler’s intentions are paramount, and cognitively 63

demanding flexibility is often necessary to interpret meaning [33,34]. Ape gesturing 64

is the only non-human communication system with substantial evidence for 65

intentional use [6-19]; providing a unique opportunity to examine the meanings, 66

analogous to human linguistic meanings, of non-human signals. Ape gestures show at 67

least first order intentionality: they are produced with the purpose of changing the 68

recipient’s behavior [35]. We present a systematic analysis of meaning for the 69

gestures employed by a wild chimpanzee community. To date, the widely described 70

flexibility of gestures has been reported in terms of the variety of ‘contexts’ in which 71

a gesture is observed [8,36]. While this method avoids potential pitfalls of attempting 72

to interpret mental states of another species, it risks exaggerating flexibility where 73

gestures with a single meaning are employed across multiple contexts. One previous 74

study examined the effect on recipients of four hand gestures, concluding that 75

responses were not dependent on situational context and were ‘primarily used for 76

directing a recipient’s movement or attention’ [38]. Here we investigate 77

communication in a natural group across the full range of chimpanzee behavior; and 78

we are able for the first time to distinguish ‘real-world usage’ from the play-based 79

communication that dominates in captivity. We examine what each gesture is for: if a 80

5

gesture is used to alter the behavior of a recipient towards a specific goal, what was 81

that goal? To find out, we adopt a holistic approach to the study of meaning that uses 82

the behavior of both signaler and recipient [39], first piloted with captive groups 83

[10,14]. We therefore focus on whether a recipient’s reaction satisfied the signaler, so 84

indicating their intended meaning. An outcome that resulted in the cessation of 85

communication, and that represented a plausible desire on the part of the signaler (e.g. 86

not an aversive experience), was taken to have satisfied the signaler and termed an 87

Apparently Satisfactory Outcome (ASO; see SI). 88

89

What do chimpanzees gesture to achieve? 90

We observed 4531 gestures within 3419 bouts of intentional communication; 91

3175 bouts (4247 gestures) apparently satisfied the signaler (communication ceased 92

following the audience’s response; Table S1). We used ASOs to indicate the 93

signalers’ intended meanings; recorded ASOs were of 19 different kinds. Most ASOs 94

(17) were requests to encourage interactions to start (e.g. ‘groom me’) or to develop 95

(‘move closer’, ‘play continue’); however, two that discouraged further social 96

interaction (‘stop that’ and ‘move away’) were used broadly across contexts to negate 97

a wide range of behavior. 98

Although we identified 19 ASOs and the chimpanzee has a repertoire of at 99

least 66 gesture types [18], some gestures may have more than one meaning. In fact, 100

only 10 of the 66 gestures were used for just a single ASO, and of these 7 were 101

recorded on ≤3 occasions. The majority of the repertoire was used for multiple ASOs 102

(number of ASOs per gesture type: mean=4.6 ±3.0, mode=2, range 1-12). The extent 103

of this multiplicity or ambiguity of meaning is likely underestimated, since the 104

number of cases of a gesture type correlated positively with the number of ASOs with 105

6

which it was associated (gestures with ≥3 cases, Pearson’s correlation: r=0.75, n=43, 106

p<0.0001). However, some of these ASOs occurred at very low frequencies, raising 107

the possibility that, rather than implying genuine ambiguity, they might stem from 108

observer error, or misunderstandings by the recipient uncorrected in further 109

communication by the signaler. Eliminating those ASOs with less than 3 instances per 110

gesture type across the population as potential errors, the majority of the gestural 111

repertoire was associated with two or three meanings (mean 2.8 ASOs per gesture). 112

Moreover, in most cases (57 of the 66 gestures) at least one ASO was play-related, 113

e.g. ‘play start’. The generality with which play-related meanings occurred indicates 114

that there may be something special about play-signals. Play is the most common 115

context for gestural communication [7,10], but in play gestures are not necessarily 116

used with their normal meaning and the outcome may not reliably signal the gesture’s 117

meaning in any other context. In subsequent analyses we therefore exclude data from 118

play bouts to avoid masking the ‘real-world’ meaning of gestures (an analysis 119

including play data is provided in SI). 120

121

Do gestures have specific meanings? 122

We examined whether different gestures were associated with a specific 123

pattern of outcomes, differing from the general distribution of ASOs in gestural 124

communication. Fifteen gesture types met the conditions for inclusion in the initial 125

analysis (SI Procedures_d), and 46 individuals contributed data. We found a 126

significant effect of gesture type on distribution of ASOs (gesture: f=2.30, df=14,101 127

p=0.009, 2-way ANOVA). Thus, the frequency with which gesture types are used, 128

outside of play, towards particular ASOs varies between gesture types: gestures have 129

specific meanings. 130

7

131

Does gesture meaning vary with the identity of the signaler? 132

The appearance of multiple meanings for a single gesture might be the result 133

of variation among signalers in the ways in which they employ their gestural 134

repertoire. We therefore examined whether meaning varied with signaler identity. 135

Fifteen gesture types met the conditions for inclusion in the detailed analysis (SI 136

Procedures_d), and 46 individuals contributed data. The possible effect of individual 137

identity was examined in two ways, graphical and statistical. For each gesture type, 138

we plotted the deviation from normal distribution of the ASO distribution (as used in 139

the ANOVAs above), per individual signaler. ASOs with similar meanings are plotted 140

adjacent to one another, allowing us to distinguish visually between gestures with 141

multiple meanings that are unambiguously different (e.g. Big Loud Scratch: ‘groom 142

me’ and ‘travel with me’), and those that are more ambiguous, with several similar 143

meanings (e.g. Object shake: ‘sexual attention male’, ‘follow me’, ‘travel with me’, 144

‘move away’ etc.). These plots gave a graphical indication of whether individual 145

signalers used the same gesture in the same way (Figure S1). An additional 21 146

gestures were regularly used outside of play, but were not recorded with sufficient 147

frequency from enough individuals for parametric analysis; for these gestures, similar 148

plots, indicating whether or not signalers employed these gestures towards the same 149

distribution of ASOs, are provided (Figure S1). 150

In two gestures, Leaf clipping and Present climb on, all usage by all 151

individuals tested was exclusively for their primary ASO (Table S2). In a further three 152

gestures, Big loud scratch, Hand fling and Present groom, the primary ASO was 153

recorded significantly more often than all other (14) ASOs combined, indicating a 154

close association with the primary ASO. In one gesture, Mouth stroke, all usage by all 155

8

individuals was exclusively for the primary and secondary ASOs combined. In a 156

further three gestures, Directed push, Present sexual and Reach, the primary and 157

secondary ASOs were recorded significantly more often than all other ASOs 158

combined. In three gestures, Embrace, Object move and Object shake, the combined 159

frequency with which the primary, secondary and tertiary ASOs were recorded across 160

individuals was significantly greater than all other ASOs combined. Thus, in 12 cases, 161

of a possible 15, there was statistical evidence of an association across individuals 162

with particular outcomes. 163

We found a statistical effect of individual identity in only two of the 15 164

gesture types: Hand on and Touch other (Table S2), with a borderline effect (p=0.058) 165

in Slap object. Both Hand on and Touch other have a clear primary function shared 166

across individuals (respectively: ‘stop that’ and ‘acquire object’); however, their 167

secondary functions varied between individuals, although with common themes of 168

social interaction or negotiation (‘move closer’, ‘move away’, ‘climb on me’, ‘climb 169

on you’). Thus, while some gestures have ambiguous meaning, the majority does not; 170

and gestures used for specific meanings are primarily used in the same way across 171

individuals. 172

173

What does each gesture mean? 174

Having shown that gestures are employed for specific outcomes by all 175

individuals, we next examined gesture meaning(s). Thirty-six gestures were suitable 176

for the analysis of their ASO distributions in contexts other than play; the gestures 177

associated with each ASO as both a primary outcome and secondary outcome are 178

listed in Table 1. In 35 of the 36 cases there was a significant association between 179

gesture type and ASO distribution (Table S3: details of analysis in SI). 180

9

As almost all gestures (32/36) were used towards more than one ASO, we 181

sought a convenient way of describing their level of ambiguity in meaning. Following 182

Cartmill & Byrne [13], we took gestures used towards a single ASO 70% of the time 183

to have ‘tight meanings’, while gestures used 50-70% of the time towards a single 184

ASO were considered to have ‘loose meaning’; all other cases were considered to be 185

‘ambiguous’. On this basis, 13 gesture types had tight meaning, 11 loose meaning, 186

and 12 were ambiguous. 187

188

Which outcomes are associated with the most gesture types? 189

Thirty-six gestures were associated with 13 non-play ASOs as either a primary 190

or secondary outcome (Tables 2 & S3). We recorded how many times a particular 191

ASO was recorded as being the primary, secondary or tertiary meaning of a gesture 192

type. ASOs varied in the number of gestures for which they were a primary outcome, 193

between 0 and 9 gestures, and for which they were a primary or secondary outcome, 194

between 0 and 16 gestures. In rank ordering, the pattern is the same, whether primary 195

alone or both primary and secondary meanings were assessed (Table 2). The number 196

of gesture types associated with an ASO might be an effect of sample size, i.e. rarely 197

observed outcomes are recorded less often and have fewer gesture types associated 198

with them; however, that was not the case here. Neither the number of gestures 199

associated with an ASO as their primary outcome (Pearson’s correlation: r=0.38, 200

n=15, p=0.16), nor primary and secondary outcomes combined (Pearson’s correlation: 201

r=0.34, n=15, p=0.22), were correlated to the number of cases of that ASO. 202

203

Discussion 204

205

10

Chimpanzees use their gestures in purposeful communication with other 206

chimpanzees; as such they can be considered meaningful [32]. In the current study of 207

wild chimpanzees, living under conditions that permit the complete expression of 208

their natural behavior, we analyzed the meanings of 36 gestures, finding them used 209

intentionally to achieve 15 purposes, other than in play. There was considerable 210

similarity of use across individuals, indicating that meanings are inherent to gestures, 211

as opposed to idiosyncratic to particular individuals or subgroups of individuals. 212

Similar indications of specific meaning were found in studies of captive 213

orangutan and gorilla gesturing [10,14]. However, in those studies no analysis of 214

individual differences was possible, and gestures used in play were included in 215

analyses (a necessarily consequence of the limited range of behavior expressed in 216

captive groups). Any analysis of meaning from data sets including play should be 217

interpreted with caution. Although playful usage should not be confounded with real-218

world usage in the analysis of meaning, play may serve as an important learning 219

environment for communication. Play allows younger individuals a safe testing 220

ground for their exploratory use of gesture, towards potentially risky goals such as 221

sexual solicitation or social negotiation. Our method of deciding intended meaning 222

works well in non-play contexts, whereas if data from play are included the 223

overwhelming dominance of play within the overall data set can ‘swamp’ any real 224

statistical association between gesture and (non-play) meaning. 225

Setting aside playful uses greatly reduces the apparent ‘ambiguity’ of gesture 226

meanings: 35 of the 36 gestures have specific individual patterns of meaning used 227

towards 1-3 of the 15 intended outcomes. The degree of ambiguity remaining is not 228

uniform across the repertoire. Some gestures are unambiguous, employed consistently 229

towards a single meaning, for example Leaf Clipping is used only to acquire ‘sexual 230

11

attention’. Others appear ambiguous: for example Grab is used for ‘stop that’ and 231

‘climb on me’ and ‘move away’ etc. (Figure S1). Appearance of ambiguity may arise 232

in part from the difficulty for human observers in discerning subtle variations in the 233

nature of the contact. It is evident to a human recipient whether or not a gentle touch 234

is intended to reposition us or to prevent us from moving; however, those distinctions 235

are visually very difficult to distinguish. Finally, gestures can be employed towards 236

two or three outcomes of a very similar nature, for example Push is used for both 237

‘move away’ and ‘stop that’. This last category is perhaps most similar to the type of 238

broad semantic class of information expressed in primate vocalizations, where an 239

alarm call rarely indicates (say) a leopard, specifically, but is rather used towards a 240

range of similar ground-based threats [28]. 241

We found considerable variation in whether an intended meaning was signaled 242

by a single gesture type or several gestures of apparently equivalent meaning. 243

Intriguingly, the degree of this ‘redundancy’ appeared to co-vary with the need for 244

context to fine-tune intended meaning. Our method necessarily restricts analysis to 245

that of imperative demands (declarative communication requires no overt change in 246

recipient behavior to satisfy the signaler). However, amongst these imperative 247

demands we could distinguish different types of meaning, co-varying with the number 248

of gestures used to express it. 249

Where we found that an intended meaning was conveyed by several different 250

gestures, the desired outcome was often apparently one that required some negotiation 251

or persuasion. For example, a request to give affiliative ‘contact’ (Embrace, Rump 252

rub, Shake hands, Bite) does not have a canonical form of response that is always 253

appropriate: exactly what the signaler wants by giving the gesture may often only 254

become clear after some further interaction. In contrast, meanings typically conveyed 255

12

by a single gesture were often well-defined and unitary: for example, ‘initiate 256

grooming’ (Big Loud Scratch). 257

The subtle regulation of individual social relationships is critical to 258

chimpanzee reproductive strategy, in which strong alliances are formed with related 259

or un-related individuals of both sexes. These relationships can impact on mating 260

success, contributing towards individual fitness. Interpretation will be aided by the 261

integration of contextual cues, some of which may be quite subtle (an individual 262

starting to move in a certain direction, or prior experiences of interacting with a 263

particular signaler). We suggest that in addition the availability of multiple gestures 264

for meanings involved in social negotiation allows for equally subtle distinctions: 265

allowing for room to maneuver in negotiation of outcomes. The majority of non-play 266

use, of the gesture types that are generally employed in play, was in social negotiation 267

meanings, such as ‘follow me’, or ‘move away’. It may be that play is used to explore 268

socially delicate communications: even though gesture meanings are basically 269

species-typical, a young ape may have much to learn about the appropriateness of 270

using gestures in particular social contexts. 271

272

273

Procedures 274

275

Observations were made on chimpanzees within the Sonso community during three 276

field periods between October 2007 and August 2009. We used focal behavior 277

sampling [40], and filmed all recorded cases of gestural communication using a Sony 278

Handycam. We defined gestures as discrete, mechanically ineffective physical 279

movements of the body observed during intentional communication (see [18]). 280

13

Movements of the whole body, limbs, and head were included; but not facial 281

expressions or static body postures. Following Call and Tomasello [8; see also 18], 282

we restricted analysis to only those gestures for which there was evidence that they 283

were used intentionally, in a goal-directed way. 284

For each individual, for each gesture type, we recorded the frequency of each 285

ASO. To remove any effect of pseudo-replication, these data were converted to 286

proportions for each individual. Thus, we calculated the proportion of the total 287

number of uses, by that individual, of that gesture, that corresponded to each ASO. 288

Then, in order to identify reliable differences in usage between gestures, we 289

calculated the overall proportion of each ASO in the data set, pooled across all other 290

gesture types, for each individual; and this ‘general distribution of ASOs’ served as a 291

null against which the actual distribution for any particular gesture type could be 292

compared. See SI Procedures for full details of all analyses. 293

All research reported was approved by the School of Psychology under the 294

approval of the University of St Andrews Animal Welfare and Ethics Committee. 295

296

297

Acknowledgements 298

299

We thank all the staff of the Budongo Conservation Field Station, the BCFS project’s 300

founder Vernon Reynolds and its current scientific director Klaus Zuberbühler. For 301

permission to work in Uganda we thank the Uganda National Council for Science and 302

Technology, the President’s Office, the Uganda Wildlife Authority and the National 303

Forestry Authority. Fieldwork of CH was generously supported by grants from 304

Wenner-Gren Foundation and Russell Trust. We thank Richard Moore and four 305

14

anonymous referees for their helpful comments on a previous version of the 306

manuscript. 307

308

15

References 309

1. Goodall, J. (1968). The Behavior of Free-living Chimpanzees in the Gombe stream 310

reserve. Anim. Behav. Mono. 1, 163-311. 311

2. Plooij, F.X. (1978). Some basic traits of language in wild chimpanzees. In Action, 312

gesture and symbol: The emergence of language, A. Lock, ed. (London, UK: 313

Academic Press) pp. 111-131. 314

3. McGrew, W.C., Marchant, L.F., Scott, S. and Tutin, C.E.G. (2001). Intergroup 315

differences in a social custom of wild chimpanzees: the grooming hand-clasp of 316

the Mahale Mountains, Tanzania. Curr. Anthropol. 42(1), 148-153. 317

4. Nishida, T. (1980). The leaf-clipping display: a newly-discovered expressive 318

gesture in wild chimpanzees. J. Hum. Evol. 9(2), 117-128. 319

5. Matsumoto-Oda, A. and Tomonaga, M. (2005). ‘Intentional’ control of sound 320

production found in leaf-clipping display of Mahale chimpanzees. J. Ethol. 23(2), 321

109-112. 322

6. Tomasello, M., et al. (1985). The development of gestural communication in young 323

chimpanzees. J. Hum. Evol. 14, 175-186. 324

7. Tomasello. M., Gust, D. and Frost, G.T. (1989). A longitudinal investigation of 325

gestural communication in young chimpanzees. Primates 30, 35-50. 326

8. Tomasello, M., Call, J., Nagell, K., Olguin, R. and Carpenter, M. (1994). The 327

learning and use of gestural signals by young chimpanzees – a trans-generational 328

study. Primates 35, 137-154. 329

9. Pika, S., Liebal, K. and Tomasello, M. (2003). Gestural communication in young 330

gorillas (Gorilla gorilla): gestural repertoire, learning and use. Am. J. Primatol. 331

60, 95-111. 332

16

10. Genty, E., Breuer, T., Hobaiter, C and Byrne, R.W. (2009). Gestural 333

communication of the gorilla (Gorilla gorilla): repertoire, intentionality and 334

possible origins. Anim. Cogn. 12(3), 527-546. 335

11. Pika, S., Liebal, K. and Tomasello, M. (2005). Gestural communication in 336

subadult bonobos (Pan paniscus): repertoire and use. Am. J. Primatol. 65(1), 39-337

61. 338

12. Liebal, K., Pika, S. and Tomasello, M. (2006). Gestural communication of 339

orangutans (Pongo pygmaeus). Gesture 6, 1-36. 340

13. Cartmill, E. and Byrne, R.W. (2007). Orangutans modify their gestural signaling 341

according to their audience’s comprehension. Curr. Biol. 17(15), 1345-1348. 342

14. Cartmill, E. and Byrne, R.W. (2010). Semantics of primate gestures: intentional 343

meaning of orangutan gestures. Anim. Cogn. 13(6), 793-804. 344

15. Pollick, A.S., Jensen, A. and DeWaal, F.B.M. (2007). Gestures and multimodal 345

signaling in bonobos. In The bonobos: behavior, ecology and conservation. 346

Furuichi, T. and Thompson, J., eds. (New York, USA: Springer). 347

16. Leavens, D.A. and Hopkins, W.D. (1998). Intentional communication by 348

chimpanzees: a cross-sectional study of the use of referential gestures. Dev. 349

Psychol. 34, 813-822. 350

17. Leavens, D.A., Russell, J.L. and Hopkins, W.D. (2005). Intentionality as 351

measured in the persistence and elaboration of communication by chimpanzees 352

(Pan troglodytes). Child Dev. 76, 291-306. 353

18. Hobaiter, C. and Byrne, R.W. (2011). The gestural repertoire of the wild 354

chimpanzee. Anim. Cogn. 14(5), 745-767. 355

19. Hobaiter, C. and Byrne, R.W. (2011). Serial gesturing by wild chimpanzees: its 356

nature and function for communication. Anim. Cogn. 14(5), 827-838. 357

17

20. Rendall, D., Owren, M.J. and Ryan, M.J. (2009). What do animal signals mean? 358

Anim. Beh. 78(2), 233-240. 359

21. Crockford, C., Wittig, R.M., Mundry, R. and Zuberbühler, K. (2012). Wild 360

chimpanzees inform ignorant group members of danger. Curr. Bio. 22(2), 142-361

146. 362

22. Schel, A.M., Townsend, S.W., Machanda, Z., Zuberbuhler, K. and Slocombe, 363

K.E. (2013). Chimpanzee alarm call production meets key criteria for 364

intentionality. Plos One. 8(10), e76674. 365

23. Hewes, G.W. (1973). Primate Communication and the Gestural Origin of 366

Language. Curr. Anth. 33, 65-84. 367

24. Armstrong, D., Stokoe, W. and Wilcox, S. (1995). Gesture and the nature of 368

language (Cambridge, UK, Cambridge University Press). 369

25. Corballis, M. (2002). From hand to mouth: The origins of language (New Jersey, 370

USA, Princeton University Press). 371

26. Cheney, D.L. (1992). How monkeys see the world: inside the mind of another 372

species (Chicago, USA, University of Chicago Press). 373

27. Smith, W.J. (1965). Message, meaning, and context in ethology. Am. Nat. 99, 374

405-409. 375

28. Seyfarth, R.M., Cheney, D.L. and Marler, P. (1980). Vervet monkey alarm calls: 376

semantic communication in a free-ranging primate. Anim. Behav. 28, 1070-1094. 377

29. Zuberbühler, K. (2003). Referential signaling in non-human primates: cognitive 378

precursors and limitations for the evolution of language. Adv. Stud. Behav. 33, 379

265-307. 380

30. Zuberbühler, K., Cheney, D.L. and Seyfarth, R. M. (1999). Conceptual semantics 381

in a nonhuman primate. J. Comp. Psychol. 113, 33-42. 382

18

31. Seyfarth, R.M. and Cheney, D.L. (2003). Signalers and receivers in animal 383

communication. Ann. Rev. Psychol. 54(1), 145-173. 384

32. Grice, H.P. (1991). Studies in the way of words (Cambridge, MA: Harvard 385

University Press). 386

33. Scott-Phillips, T. (2010). Animal communication: insights from linguistic 387

pragmatics. Anim. Behav. 79(1), e1-e4. 388

34. Wheeler, B.C. and Fischer, J. (2012). Functionally referential signals: a promising 389

paradigm whose time has passed. Evol. Anthropol. 21, 195-205. 390

35. Dennett, D.C. (1987) The intentional stance (Cambridge, MA: MIT Press). 391

36. Liebal, K., Call, J. and Tomasello, M. (2004). Use of gesture sequences in 392

chimpanzees. Am. J. Primatol. 64(4), 377-396. 393

38. Roberts, A.I., Vick, S.J. and Buchanan-Smith, H.M. (2012). Usage and 394

comprehension of manual gestures in wild chimpanzees. Anim. Behav. 84(2), 395

459-470. 396

39. Grice, H.P. (1957). Meaning. The Philosophical Review 66(3), 377-388. 397

40. Altmann, J. (1974). Observational study of behavior: sampling methods. 398

Behavior, 49(3-4), 227-267. 399

19

Table 1: Gestural Lexicon 400 Real-world meanings are defined and listed with the gestures to which they are 401 associated, either a primary or a secondary outcome (Table S3: data from all 402 individuals, raw scores converted to proportions). Meanings are consistent across 403 individuals (Figure S1 and Table S2). Meanings are ordered in declining order of the 404 number of gestures used to effect them: note that negation (‘stop that’ and ‘move 405 away’) can be achieved with the largest variety of gestures, and that more alternatives 406 are available for social negotiation than for simple requests. 407 Meaning: definition [primary outcome of these gestures] {secondary outcome of these gestures} 408 409 410 Stop that: either cease behavior previously directed towards the signaler or change behavior to direct it towards another [primary: Grab; Hand on; Jump; Push; Side roulade; Slap other; Somersault; Stomp 2-feet; Tap other] {secondary: Arm swing; Bite; Foot present; Hand fling; Punch other; Shake hands; Slap object} Move away: move away from signaler [primary: Arm swing; Hand fling; Jump; Object shake; Punch object/ground; Punch other; Slap object] {secondary: Arm raise; Object move; Push; Slap other; Stomp; Tap other} Contact: physical contact of apparently affiliative nature, e.g. hugging, touching etc. [primary: Bite; Embrace; Rump rub; Shake hands] {secondary: Present sexual; Reach; Touch other} Acquire object: give signaler object [primary: Arm raise; Mouth stroke; Reach; Touch other] {secondary: Hand on} Follow me: mature recipient follows mature signaler, usually in consortship [primary: Jump; Slap object with object; Throw object] {secondary: Foot present; Rump rub; Stomp 2-feet} Move closer: move closer [primary: Beckon; Grab-pull; Slap object with object] {secondary: Arm swing; Directed push; Mouth stroke}

Sexual attention to male: ♀-responds sexually [primary: Leaf-clipping; Object move; Stomp] {secondary: Object shake; Punch object/ground} Climb on me: climb on signaler’s body [primary: Foot present; Present climb on] {secondary: Grab; Grab-pull} Initiate grooming: grooming between signaler and recipient [primary: Big loud scratch] {secondary: Bite; Present grooming} Sexual attention to female: ♂-responds sexually [primary: Present sexual] {secondary: Leaf clipping} Reposition body: move (and hold) body into indicated position [primary: Directed push] {secondary: Beckon} Attend to specific location: adjust behavior to focus attention on indicated location [primary: Present grooming] {secondary: none} Travel with me (adult): travel together with adult signaler [primary: n/a*] {secondary: Big loud scratch; Embrace} Climb on you: permits signaler to climb on [primary: n/a*] {tertiary**: Reach} Travel with me (infant): travel together with infant signaler [primary: n/a*] {other**: Big loud scratch; Grab-pull; Poke}

* only recorded as secondary outcome ** rarely observed outcome; only recorded as tertiary or less frequent 411 412

20

Table 2: Primary or secondary gesture meanings (excluding play) 413 The ASOs (as defined in Table S1) recorded as the primary, secondary, or tertiary 414 ASO for each gesture type (Table S3). ASOs are listed in order of the number of 415 gesture types (N) associated with them as their primary, and then secondary, or 416 tertiary outcome. 417 *Both these ASOs were recorded only as the tertiary or even less frequent outcome of 418 a gesture type, as used by the community as a whole. However, their use was 419 necessarily limited to young infant signalers; evidently they would be more 420 prominently represented in a study of infant gesturing. 421 422

ASO N1 (primary) N1+2 (primary or secondary)

N1+2+3 (primary or secondary or tertiary)

Stop that 9 16 20

Move away 7 13 14

Contact 4 7 10

Acquire object 4 5 8

Follow me 3 6 10

Move closer 3 6 8

Sex attention (to male) 3 5 7

Climb on me 2 4 6

Initiate grooming 1 3 4

Sex attention (to female) 1 2 2

Reposition body 1 2 2

Attend to specific location 1 1 1

Travel with me (adult) 0 2 2

Climb on you* 0 0 1

Travel with me (infant)* 0 0 0

423 424

Figure S1: Related to Table 1 Individual use of gesture types: graphical analysis The mean percentage deviation from the normal distribution of ASOs in gestural communication is plotted per signaler for the 15 gesture types suitable for parametric analysis; followed by the 21 gestures analyzed with non-parametric analysis (see Table S3: evidence). Number of individual signalers and total number of cases of the gesture are provided. ASOs with potentially similar meaning are plotted adjacent to each other. Each line represents a distribution of usage across the ASOs; the weight of the line correlates directly to the number of signalers that employed this distribution of meaning. For clarity only ASOs for which the gesture was employed are labelled. *Note that, in the gestures analyzed with parametric analysis individual contributed several cases of the gesture: thus any single errors (either in attribution by the coder, or understanding by the recipient) represent only a portion of the individual signaler’s usage. For gestures analyzed through nonparametric analysis, an individual may contribute only a single case of the gesture so, even where this was an error it then represents 100% of their usage. Thus, errors and variation are potentially exaggerated with respect to distributions shown in Figure S1.

15 gestures analyzed through parametric analysis (Table S3) Big Loud Scratch: individuals n=18, cases n=177 Directed push: individuals n=16, cases n=157

Embrace: individuals n=4, cases n=118 Hand fling: individuals n=5, cases n=39

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Atte

nd

Gro

om

Clim

b on

me

Follo

w

Trav

el w

ith

(ad)

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

obj

ect

Rep

ositi

on

Clim

b on

me

Sex

att

(m)

Trav

el w

ith

(ad)

M

ove

clos

er

Con

tact

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Gro

om

Follo

w

Con

tact

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Mov

e aw

ay

Sto

p

Hand on: individuals n=4, cases n=21 Leaf clip: individuals n=6, cases n=27

Mouth stroke: individuals n=3, cases n=15 Object move: individuals n=3, cases n=52

Object shake: individuals n=13, cases n=303 Present climb on: individuals n=3, cases n=17

Present grooming: individuals n=27, cases n=156 Present sexual: individuals n=18, cases n=135

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Mov

e cl

oser

Mov

e aw

ay

Sto

p -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sex

att

(m)

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

ob

ject

Mov

e cl

oser

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sex

att

(m)

Follo

w

Mov

e cl

oser

Mov

e aw

ay

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

obj

ect

Gro

om

Sex

att

(m)

Follo

w

Trav

el w

ith

Mov

e cl

oser

C

onta

ct

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Clim

b on

me

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Atte

nd

Gro

om

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

obj

ect

Atte

nd

Sex

atte

ntio

n (f)

Follo

w

Con

tact

Sto

p

Reach: individuals n=25, cases n=228 Slap object: individuals n=7, cases n=48

Touch other: individuals n=9, cases n=42

21 gesture analyzed through non-parametric analysis (Table S3)* Arm raise: individuals n=12, cases n=27 Arm swing: individuals n=10, cases n=39

Beckon: individuals n=4, cases n=6 Bite: individuals n=6, cases n=7

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

obj

ect

Atte

nd

Clim

b on

me

Clim

b on

you

Sex

ual a

tt (m

)

Mov

e cl

oser

C

onta

ct

Mov

e aw

ay

Sto

p -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sex

att

(m)

Follo

w

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

obj

ect

Clim

b on

me

Clim

b on

you

Trav

el w

ith (a

d)

Mov

e cl

oser

C

onta

ct

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

ob

ject

Clim

b on

you

Mov

e cl

oser

Mov

e aw

ay

Sto

p -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

ob

ject

Follo

w

Mov

e cl

oser

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Rep

ositi

on

Mov

e cl

oser

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Gro

om

Con

tact

Sto

p

Foot present: individuals n=4, cases n=8 Grab: individuals n=13, cases n=19

Grab-pull: individuals n=8, cases n=16 Jump: individuals n=3, cases n=18

Slap object with object: individuals n=2, cases n=15

Slap other: individuals n=14, cases n=17

Somersault: individuals n=6, cases n=15 Stomp: individuals n=12, cases n=30

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Clim

b on

me

Follo

w

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Atte

nd

Rep

ositi

on

Clim

b on

me

Clim

b on

yo

u

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Clim

b on

me

Clim

b on

you

Mov

e cl

oser

C

onta

ct

Trav

el w

ith

(in)

Mov

e aw

ay

Sto

p -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Atte

nd

Gro

om

Sex

att

(m)

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Follo

w

Mov

e cl

oser

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sex

att

(m)

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Sex

att

(m)

Follo

w

Mov

e cl

oser

Mov

e aw

ay

Sto

p

Stomp 2-feet: individuals n=5, cases n=10 Tap other: individuals n=12, cases n=15

Throw object: individuals n=1, cases n=5*

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Follo

w

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Acq

uire

ob

ject

Mov

e cl

oser

C

onta

ct

Mov

e aw

ay

Sto

p

-0.2 0.0 0.2 0.4 0.6 0.8 1.0

Follo

w

Table S1: Related to Table 2 Apparently Satisfactory Outcomes in chimpanzee gestural communication. Definitions of ASOs are provided along with the percentage frequencies with which single gestures or bouts of gesturing were employed towards a particular ASO; raw frequencies of the number of gesture instances are given in brackets (n). *Note that the ‘Unknown’ category does not accurately reflect the rate of failed communication attempts: since the definition of a gesture requires evidence of intentional usage, many genuinely communicative instances were liable to be missed, and these may disproportionately include cases where the target audience does not react. ASO Definition Gestures

% (n) Bouts % (n)

Acquire object recipient gives signaler object (e.g. food, leaf sponge, etc.)

6.3% (283)

4.8% (164)

Attend to specific location

recipient adjusts their behavior to focus attention on the location indicated in the signaler’s gestural communication, usually in grooming

4.0% (183) 5.0% (172)

Climb on me recipient climbs on signaler’s body 1.8% (83) 2.1% (72) Climb on you recipient permits signaler to climb on them 0.3% (15) 0.4% (13) Contact physical contact of an apparently affiliative

nature, such as hugging, touching etc. between the signaler and recipient

2.9% (132) 3.5% (119)

Follow me mature recipient follows mature signaler, usually in consortship

8.2% (373) 4.2% (145)

Initiate grooming grooming between the signaler and recipient 3.8% (174) 4.2% (143) Move away recipient moves away from signaler 3.8% (170) 3.8% (129) Move closer recipient moves closer to signaler 2.4% (107) 2.2% (75) Play start recipient starts to play with the signaler 42.7% (1933) 43.4% (1485) Play change: increase intensity

recipient changes the type of play from chasing play to contact (e.g. wrestling) play

0.7% (33) 0.8% (26)

Play change: decrease intensity

recipient changes the type of play from contact to chasing play

0.1% (3) 0.1% (3)

Play resume recipient resumes playing after a pause in the activity

5.7% (257) 6.3% (214)

Reposition body the recipient moves (and holds) their body into the indicated position

2.2% (100) 2.7% (93)

Sexual attention to female

(male) recipient responds sexually to signaler attention (e.g. inspection, copulation etc.)

2.4% (107) 2.0% (69)

Sexual attention to male

(female) recipient responds sexually to signaler (e.g. presentation, copulation, etc.)

1.7% (110) 3.1% (107)

Stop that the recipient either ceases behavior previously directed towards the signaler or changes their behavior to direct it towards another individual

3.2% (145) 3.7% (127)

Travel with me (adult) recipient travels together with adult signaler 0.8% (36) 0.5% (18) Travel with me (infant) recipient travels together with infant signaler 0.1% (3) 0.1% (1) Unknown* intentional gesture fails to elicit any response, so

unable to attribute the intended purpose 6.3% (284) 7.1% (244)

Table S2: Related to Table 1 Individual use of gesture types: statistical analysis. A significant result indicates that the gesture has a distinct meaning, which does not vary with signaler identity. A significant t-test or G-test result indicates a population-general effect: that is, that there is no significant effect of signaler identity on the pattern of use of an individual gesture type (for its primary; primary and secondary; or primary, secondary and tertiary ASOs combined). ASOs were derived from the analysis of data from all individuals, shown in table S3; tests were run on the subset of individuals who contributed at least 3 cases of gesture use, as in the parametric analyses in table S3. *All cases of gesture use tested were towards the primary ASO

Gesture ASOs: primary (secondary)

[tertiary] ASOs tested

Evidence

Big loud scratch

Initiate grooming 1 T=2.88, df=17, p=0.010

Directed push Reposition body (Move closer) 1,2 T=3.26, df=15, p=0.005 Embrace Contact (Travel with me adult)

[Follow me] 1,2,3 T=7.00, df=3, p=0.006

Hand fling Move away (Stop that) 1 T=11.00, df=4, p<0.001 Hand on Stop that (Acquire object)

[Contact] 1,2,3 T=0.46, df=3, p=0.680

Leaf clipping Sexual attention to male* 1 G=37.4, df=1, p<0.001 Mouth stroke Acquire object (Move closer) 1,2 G=20.8, df=1, p=0.003 Object move Sexual attention to male

(Move away) [Follow] 1,2,3 T=13.46, df=3, p=0.005

Object shake Move away (Sexual attention to male) [Acquire object]

1,2,3 T=3.21, df=12, p=0.008

Present climb on

Climb on me* 1 G=23.6, df=1, p<0.001

Present grooming

Attend to specific location 1 T=34.96, df=26, p<0.0001

Present sexual Sexual attention to female (Contact)

1,2 T=14.20, df=17, p<0.0001

Reach Acquire object (Contact) 1,2 T=3.134, df=24, p=0.005 Slap object Move away (Stop that) [Sexual

attention to male] 1,2,3 T=2.34, df=6, p=0.058

Touch other Acquire object (Contact) [Stop that]

1,2,3 T=1.02, df=8, p=0.338

Table S3: Related to Tables 1 and 2 Gesture meanings. Most commonly recorded ASOs, for each gesture type. Percentages are calculated from cases of non-play gestural communication and do not include cases assigned an ASO of ‘unknown’. Only the commonest two ASOs recorded are shown; data are from all individuals, with each individual’s raw score converted to a proportion. We carried out individual ANOVAs or Chi-square tests as appropriate for each of the 36 gestures suitable for analysis, comparing the ASO distribution for each gesture against the null distribution of the ASOs in the data set as a whole (see SI Procedure). In addition we attempt to identify gesture types specifically employed in play communication; we analyze all usage (including play) for the 18 gestures where insufficient evidence of meaning was available from non-play contexts. Statistical evidence is provided for any interaction between gesture type and distribution of ASOs, tested using either an ANOVA, or, where insufficient cases for parametric analysis, Chi-square with individual data pooled. A significant finding indicates that the distribution of ASOs for the specific gesture types varies significantly from the null distribution of ASOs. N is the number of individuals contributing data to the test, which is the number of subjects tested when the statistical test is an ANOVA in the adjacent ‘evidence’ column; note that for parametric tests only individuals contributing at least 3 cases of gesture use were included (see SI Procedure_d). The total number of cases of gestures used (n) is also provided, and this is the number of subjects tested when the adjacent statistical test is a Chi-square. Individual identity was coded and treated as a random variable in all ANOVAs. * It was not possible to control for individual identity in Chi-square tests; thus, p values of chi-square tests should be interpreted with caution. See SI Procedure_d for discussion of the interpretation of pooled data. ** Throw object was recorded 100% of the time towards a single outcome (‘follow me’), but with only 5 cases the interaction only approached significance (p=0.08). *** Note that in the case of gestures analyzed with play data, because of the prevalence of play data within the total (null) data set, a failure to show deviation from the null is here much more likely. While a significant result will indicate a specific pattern of use for that gesture type, a non-significant result does not necessarily indicate that there is no specific pattern of use: rather it gives no information in either direction. While 17 of the 18 gestures had a tight meaning of ‘play start’, the extremely high frequency with which play was recorded as an outcome in the dataset meant that for 8 gestures the distribution of ASO did not vary from the overall average distribution of ASOs. While this does not represent a significant proportion of the repertoire (Binomial: n=18, p=0.815), the non-significance is likely due to the overwhelming predominance of play data in the set, and should only be interpreted as providing no specific evidence, rather than evidence against meaningful use of gestures used in play.

Gesture meanings: non-play gesture type ASO (%) all individuals N (n) evidence*

Arm raise Acquire object 38% (Move away 29%) 12 (27) χ2=65.71, df=14 p<0.0001

Arm swing Move away 40% (Move closer/Stop that 23%) 10 (39) χ2=62.64, df=14 p<0.0001

Beckon Move closer 63% (Reposition body 38%) 4 (6) χ2=52.30, df=14 p<0.0001

Big loud scratch Initiate grooming 82% (Travel with me - adult 9%)

18 (177) f=45.33, df=14, 238 p<0.001

Bite Contact 67% (Stop that/Initiate grooming 17%) 6 (7) χ2=37.76, df=14 p=0.0006

Directed push Reposition body 57% (Move closer 15%) 16 (157) f=29.13, df=14, 210 p<0.001

Embrace Contact 89% (Travel with me – adult 7%) 4 (18) f=33.83, df=14,42 p<0.001

Foot present Climb on me 50% (Follow me/Stop that: 25%) 4 (8) χ2=104.17, df=14 p<0.0001

Grab Stop that 44% (Climb on me 18%) 13 (19) χ2=123.57, df=14 p<0.0001

Grab-pull Move closer 35% (Climb on me 27%) 8 (16) χ2=117.36, df=14 p<0.0001

Hand fling Move away 73% (Stop that 27%) 5 (39) f=427.5, df=14,56 p<0.001

Hand on Stop that 50% (Acquire object 25%) 4 (21) f=3.28, df=14,42 p=0.001

Jump Follow me/Stop that/Move away 33% 3 (18) χ2=48.47, df=14 p<0.0001

Leaf clipping Sexual attention to male 89% (Sexual attention to female 11%)

6 (27) f=1925.79, df=14,70 p<0.001

Mouth stroke Acquire object 93% (Move closer 7%) 3 (15) f=4.18, df=14,28 p=0.001

Object move Sexual attention to male 45% (Move away 26%) 3 (52) f=2.16, df=14,28 p=0.041

Object shake Move away 37% (Sexual attention to male 35%) 13 (303) f=7.68, df=14,168 p<0.001

Present (climb on) Climb on me 100% 3 (17) f=1820.37, df=14,28 p<0.001

Present (grooming) Attend to specific location 99% (Initiate grooming 1%)

27 (156) f=2384.95, df=14,364 p<0.001

Present (sexual) Sexual attention to female 49% (Contact 33%) 18 (135) f=50.80, df=14,238 p<0.001

Punch object/ground

Move away 42% (Sexual attention to male 33%) 6 (8) χ2=24.94, df=14 p=0.0352

Punch other Move away 57% (Stop that 29%) 7 (7) χ2=29.17, df=14 p=0.0099

Push Stop that 78% (Move away 22%) 12 (18) χ2=143.82, df=14 p<0.0001

Reach Acquire object 53% (Contact 20%) 25 (228) f=30.24, df=14,336 p<0.001

Rump rub Contact 80% (Follow 20%) 5 (5) χ2=45.31, df=14 p<0.0001

Shake hands Contact 83% (Stop that 17%) 6 (8) χ2=68.18, df=14 p<0.0001

Side roulade Stop that 100% 4 (9) χ2=116.00, df=14 p<0.0001

Slap object Move away 60% (Stop that 13%) 7 (48) f=10.44, df=14,84 p<0.001

Slap object with Follow me/Move closer 50% 2 (15) χ2=56.89, df=14 p<0.0001

object

Slap other Stop that 64% (Move away 32%) 14 (17) χ2=100.47, df=14 p<0.0001

Somersault Stop that 100% 6 (15) χ2=193.33, df=14 p<0.0001

Stomp Sexual attention to male 38% (Move away 29%) 12 (30) χ2=48.83, df=14 p<0.0001

Stomp 2-feet Stop that 50% (Follow me 40%) 5 (10) χ2=25.63, df=14 p=0.0288

Tap other Stop that 42% (Move away 25%) 12 (15) χ2=42.00, df=14 p<0.0001

Throw object ** Follow me 100% 1 (5) χ2=22.03, df=14 p=0.0781

Gesture meanings: play*** Arm shake Play start 81% (Follow me 8%) 3 (13) f=8.01, df=18,36 p<0.001

Arm wave Play start 80% (Follow me 20%) 5 (5) χ2=4.31, df=18 p=0.999

Dangle Play start 86% (Play resume 14%) 17 (280) f=388.05, df=18,288 p<0.001

Drum object palms Play start 81% (Sexual attention to male 19%) 7 (10) χ2=20.07, df=18 p=0.329

Feet shake Play start 70% (Play resume 26%) 9 (14) χ2=21.34, df=18 p=0.263

Gallop Play start 85% (Move away/ Play change - decrease intensity: 7%)

5 (28) f=394.3, df=18,72 p<0.001

Hand shake Play start 82% (Play change - increase intensity 9%, Contact 9%)

11 (11) χ2=20.30, df=18 p=0.32

Head nod Play start 78% (Play resume 12%) 3 (13) f=134.36, df=18,36 p<0.001

Head stand Play start 82% (Play resume 18%) 7 (32) f=90.45, df=18,108 p<0.001

Kick Play start 74% (Play resume 11%) 3 (13) f=2.12, df=18,36 p=0.027

Knock object Play start 83% (Follow me 17%) 6 (6) χ2=5.05, df=18 p=0.999

Leg swing Play start 78% (Follow me 17%) 6 (15) χ2=7.88, df=18 p=0.980

Object in mouth approach

Play start 81% (Play resume 19%) 3 (13) f=14.92 df=18,36 p<0.001

Poke Play start 72% (Play resume/Travel with me-infant: 11%)

9 (15) χ2=76.51, df=18 p<0.001

Pounce Play start 83% (Play resume 8%) 6 (7) χ2=5.94, df=18 p=0.997

Roll over Play start 68% (Play resume 16%) 5 (28) f=568.00 df=18,72 p<0.001

Stomp other Play start 100% 6 (6) χ2=7.19, df=18 p=0.989

Stomp other 2-feet Play start 93% (Play resume 7%) 3 (13) f=52.32, df=18,36 p<0.001

SUPPLEMENTAL PROCEDURES

(a) Study site and subjects

The Budongo Conservation Field Station was established in 1990 in the Budongo

Forest Reserve, which lies in the western Rift Valley in Uganda (1°350 – 1°550 N,

31°180 – 31°420 E) at a mean altitude of 1,050 m. The 793-km2 Reserve includes 482

km2 of continuous medium-altitude semi-deciduous forest cover [41]. The forest

within this site is predominantly secondary forest growth, due to regular logging until

1990.

Observations were made on chimpanzees within the Sonso community during three

field periods between October 2007 and August 2009 (October 2007–March 2008;

June 2008–January 2009; May 2009–August 2009). At the start of data collection in

October 2007, the Sonso study community of chimpanzees consisted of 81 named

individuals. Following Reynolds [42] we defined age groups as follows: infants (0–4

years), juveniles (5–9 years), sub-adults (m: 10–15 years, f: 10–14 years) and adults

(m: 16+ years, f: 15+ years). Using these categories, the initial group composition was

32 adults (7 males and 25 females), 16 sub-adults (10 males and 6 females), 15

juveniles (6 males and 9 females) and 18 infants (3 males and 15 females). Over the

course of the 22-month study, there were 10 deaths or long-term disappearances, 6

immigrations and 5 births, leaving the final total at 82.

(b) Analysis of gestures

Parametric analyses were carried out in SPSS v11, non-parametric analyses were

calculated by hand; α = 0.05 was required for significance. Means are given ±

standard deviation, throughout. All statistical tests were two-tailed.

We maintained a record of the frequency with which a particular individual

was observed: where we could choose which of several social interactions to film, we

targeted individuals previously sampled infrequently (see [18] for detailed data

collection and analysis protocol). Videos were uploaded, scanned for potential cases

of gestural communication and these cases edited into discrete clips; any gestures

were then reviewed in slow motion and coded into a Filemaker Pro Database. A

gesture sequence refers to multiple gestures given without pause i.e. <1sec between

gestures [19]. A gesture bout refers to a series of gestures or gesture sequences,

separated by ≥1sec [19], but given by the same signaler and apparently directed at

achieving the same outcome.

Only gestures made with evidence of intentional usage were analyzed. Thus,

all gesture instances considered in this paper showed one or more of the following:

audience checking, where the signaler shows awareness of the recipients and their

state of attention; response waiting, where the signaler pauses at the end of the

communication and continues visual monitoring of recipients while waiting, without

producing any further self- or other-directed activity; or persistence, where - after

response waiting, and in the absence of a response that in other cases is taken as

satisfactory - further gestures are given, apparently towards the same end. We

interpret these gestures as having at least 1st order intentionality [35]. By this we mean

that a signaler gestures with the intention of changing the recipient’s behavior.

(c) Interpretation of meaning in chimpanzee gestures

The aim of this study was to discover, for each gesture type, the purpose that a

chimpanzee intended to achieve by using it: the intended meaning. We followed the

method originally established in Genty et al. [10], of deducing the signaler’s intended

meaning from whatever action taken by the target that appears to satisfy the gesturer:

i.e. the target’s action that immediately precedes the cessation of communication by

the signaler. We refer to this as the apparently satisfactory outcome or ASO, to avoid

any implication that we had access to internal mental states.

We considered persistence in communication (i.e. continuing to signal after a

pause of ≥1s.) to imply the failure of the prior single gesture or sequence, leading to

the production of a new gesture or sequence directed towards the same ASO. In this

case, the ASO that finally resulted in cessation of communication was taken to apply

also to earlier, failed gestures or sequences within the same bout of signaling. To

contribute to identifying a gesture’s meaning, we also required the recipient’s

behavior to meet some plausible biological function for the signaler. Given the

growing body of evidence against the existence of ape gestures that serve simply to

get the recipient’s attention [10,18,19,36], we treated as communicative failures those

cases where the recipient changed their attentional state but made no further response.

Also treated as failures were cases where the recipient: left, if this did not serve any

plausible biological advantage to the signaler; physically prevented the signaler from

continuing to gesture; aggressively chased the signaler away; or showed no change in

behavior or demeanor. We suspected that gestures given when pant-grunting, a

submissive vocal signal reliably directed to dominant individuals, might in effect be

requests for the high-rank chimpanzee to remain in the same state (rather than show

aggression to the subordinate). If so, then a lack of response would be the satisfactory

outcome for the signaler: but we could find no way of distinguishing these cases

objectively from others where lack of response was simply a matter of failure. We

therefore assigned all such cases, along with all apparent failures of communication,

to the ASO category ‘unknown’.

Our method has limitations. It is limited to the description of imperative

demands that require a behavioral change in the recipient; further meanings may be

possible in chimpanzee communication, but our approach would not identify them.

By using ASOs, rather than charting the context of use of a gesture, we have avoided

the spurious ‘flexibility’ of cases in which a gesture with a single meaning is

employed across multiple contexts, for example we identify “stop that” as a single

ASO although it is employed across many contexts. However, where a signaler’s

intended meaning is compatible with a range of ASOs, our method would still tend to

exaggerate signal flexibility where several different actions might satisfy the same

intention. For example, a gestural request for ‘contact’ might be satisfied by very

different actions by the alpha male chimpanzee or by a younger sibling. These

variations might themselves be teased apart by considering the relative rank, sex, and

relationship of the signaler to the recipient. However, that would require a data set

that distinguished the many possible types of signaler-recipient relationships,

requiring a much more extensive, long-term study.

(d) Analyzing patterns of meaning: are individual gestures used towards multiple

ASOs?

A forced dependency between gesture and ASO could render the data unsuitable for

parametric analysis: however, we were able to eliminate this risk by counting all uses

of the gesture, including cases where the ASO had been assigned as ‘unknown’, in the

total number of cases, rather than just those where it was possible to define an ASO.

As expected, most gestures were associated with only one or a few ASOs: a non-

homogeneous distribution of variation. We corrected for this non-homogeneity of

variation to render the data suitable for parametric analysis (as recommended in

Snedecor and Cochran, see: [43]), as follows. In all cases where the proportion (i.e.

uses of a gesture type for any one ASO / total uses of the gesture type, n/N) was equal

to 0 or 1, we re-scaled the values. The following substitutions were used: where the

proportion was 1, the value was replaced with 1-(1/4N); where proportion was 0, the

value was replaced with 1/4N.

We compared the distribution of ASOs between all gesture types with

sufficient data for parametric analysis using a 2-way ANOVA. We required a

minimum of three different signalers, each with a minimum three cases in which an

ASO could be defined, for each gesture type in this initial ANOVA. To remove any

effect of pseudo-replication, these data were converted to proportions for each

individual. Thus, we calculated the proportion of the total number of uses, by that

individual, of that gesture, that corresponded to each ASO (see Procedures in the main

paper). As different individuals contributed varying numbers of instances of different

gesture types, and thus, the identity of the individuals contributing their data to an

analysis of a particular gesture type varied, individual identity was coded and treated

as a random factor in the analysis. We then used a planned series of separate

ANOVAs for each gesture, with ASO type as an independent variable. For the null

distribution to be an effective representation of general gesture use, the data included

in the null set should be from as many individuals and gesture types as possible.

Therefore, we used the maximum possible set of individuals (all individuals who

contributed at least 3 cases of gesture to a gesture type in the parametric analysis) to

construct the null distribution, rather than the reduced subset of individuals whose

data contributed to each specific gesture type. To address any possible problem from

the lack of matched samples, individual identity was again coded as a random

variable. In each of these ANOVAS we subtracted this null distribution from the

specific distribution per gesture type of each individual who contributed their data;

giving us a measure of the extent to which the specific distribution of ASOs for a

gesture type deviated from the normal distribution of ASOs in gestural

communication. A significant interaction between ASO type and the mean percentage

frequency deviation would signify that the distribution of ASOs for the specific

gesture type varied from the general distribution of ASOs in the data set. In any case

where insufficient data were available to run a parametric ANOVA, a Chi-square test

using raw frequencies pooled over individuals was employed, to compare the

observed frequency of ASOs for which a gesture-type was used with the ‘null’

general distribution of ASOs within the data set.

There are two possible risks in the use of chi-square tests for these analyses.

Firstly, they do not allow for us to control for any effect of individual identity. We

have already established both graphically (see first half of Figure S1) and statistically

(see Table S2) that there is no effect of individual identity on signal meaning in the

majority of the gestures tested with parametric analyses (12 of 15); and we expect the

same to hold true of the remainder of the repertoire. We offer the same graphical

support for this consistency of meaning across signalers in the second half of Figure

S1, for all gestures analyzed with non-parametric tests. Secondly, by combining data

from individuals in the non-parametric statistics, we risk artificially inflating power

and thus increase the risk of a type 1 error. However, this would only affect the extent

and not the direction of the outcome; and, again, the graphical evidence provided in

Figure S1 offers additional support for our interpretation of the meaning of the

gestures analyzed through non-parametric statistics.

While employing multiple separate tests for every gesture in the repertoire is

not an ideal method, as it runs the risk of false positives, we would still expect only a

small proportion of results to be significant through chance alone: in fact, 35 of 36

cases tested showed an association (Binomial: n=36, p<0.0001).

(e) Playful versus non-playful communication

All gestures included for analysis in this paper showed evidence of signaler

satisfaction with a recipient response, one that fulfilled a plausible biological need on

the part of the signaler, i.e. where it was possible to record an apparently satisfactory

outcome. Amongst the 19 total ASOs (Table S1), 4 were play related: ‘play start’,

‘play change: increase intensity, ‘play change: decrease intensity’, and play-resume’.

Gestures were largely analyzed as they were employed outside of play, to try to

determine their normal meaning (Table S2). Any gesture types only found in play or

recorded too infrequently outside of play for the successful attribution of an ASO

remained ambiguous: they might simply be rare over all contexts, in which case they

must remain unanalyzed; or they might genuinely function mainly in play regulation.

For these unassigned gestures, we next examined the entire data corpus, including

play. If the gesture was never, or only rarely, used outside of play, but regularly

employed within play, we assigned an ASO within play regulation; where too little

data was present even after the inclusion of the play data, the ASO remained

unassigned.

Five gesture types were seen exclusively within play and were successfully

employed 5 or more times: Feet shake, Headstand, Object in mouth approach, Stomp

2-feet other, Stomp other. A further 13 gestures were successfully employed fewer

than 5 times outside of play (and were thus not suitable for the non-play analysis in

the previous section); but with the inclusion of play data they were used 5 or more

times (Arm shake, Arm wave, Dangle, Drum object (palms), Gallop, Hand shake,

Head nod, Kick, Knock object, Leg swing, Poke, Pounce, Roll over). All of these 18

gesture types have been assigned an ASO on the basis of the data including play cases

(Table S4). For 12 gesture types we observed too few cases of clear attribution of

ASO either outside of or including play, for any type of analysis (Bow, Clap, Drum

other, Head butt, Hide face, Hit with object, Look, Pirouette, Stiff walk, Tandem

walk, Tap object, Water splash) the ASO in these cases remained unassigned.

(f) Statistical analysis of individual use of gesture types

We compared the frequency of the primary ASO (as defined for the whole community

in Table S3) with the frequency of all other ASOs combined, in a series of paired t-

tests per gesture type. Where the data were not suitable for analysis with a paired t-

test, we conducted a replicated G-test of goodness of fit, which allows for the

statistical assessment of multiple chi-squares per individual signaler after testing for

the suitability of combining them, avoiding the problem of pseudoreplication. In a

simple system we might expect that a single ASO would be employed significantly

more often than all other ASOs combined. In reality, we found that gestures are used

towards 1-3 different ASOs (see Results: What do chimpanzees gesture to achieve)

and thus any minor errors, by either the coder or the chimpanzees, are likely to have

greater impact on a smaller data set. We therefore re-tested the gesture types to check

whether or not the combination of the primary and secondary, or primary, secondary,

and tertiary, ASOs (Table S2), was observed significantly more often than the

frequency of all other ASOs combined.

SUPPLEMENTAL REFERENCES

41. Eggeling, W.J. (1947). Observations on the ecology of the Budongo rain forest,

Uganda. J. Ecol. 34, 20-87.

42. Reynolds, V. (2005). The chimpanzees of the Budongo forest, Oxford University

Press.

43. Snedecor, G.W. and Cochran, W.G. (1989). Statistical methods 7th edition (Iowa,

USA, Iowa State University Press).