Manipulable object and human contact: preferences and modulation of emotional states in weaned piglets

Enriching the life of farm animals is an obligation in intensive farming conditions. In pigs, manipulable materials are mandatory when no bedding is available. Like manipulable objects, positive human interactions might be considered as enrichment, as they provide the animals occasions to interact, increase their activity and lead to positive emotional states. In this study, we investigated how weaned piglets perceived a manipulable object, and a familiar human. After a similar familiarization to both stimuli, twenty-four weaned piglets were tested for a potential preference for one of the stimuli and submitted to isolation/reunion tests to evaluate the emotional value of the stimuli. We hypothesized that being reunited with a stimulus would attenuate the stress of social isolation and promote positive behaviors, and even more that the stimulus has a positive emotional value for piglets. Although our behavioural data did not allow to show a preference for one of the stimuli, piglets approached more often the human and were observed laying down only near the human. Using behavioural and bioacoustic data, we showed that reunion with the human decreased more the time spent in an attentive state and mobility of piglets than reunion with the object, and isolation. Vocalizations differed between reunions with the object and the human, and were different from vocalizations during isolation. The human presence led to higher frequency range, more noisy and shorter grunts. Finally, both stimuli decreased the isolation stress of piglets, and piglets seemed to be in a more positive emotional state with the human compared to the object. It confirms the potential need for positive human interactions to be used as pseudo-social enrichment in pigs.


Introduction 29
The intensive production system of animal products sometimes implies large densities of farm animals and 30 can lead to deleterious behaviors and decrease the physical or mental health of animals, i.e. their welfare. 31 Animal welfare covers, among others, the importance of the animal's ability to keep control of mental and 32 body stability in different environmental conditions (Broom 2011). Improving animal welfare is both reducing 33 negatively perceived contexts as well as increasing positively perceived contexts and species-specific 34 behaviors (Peterson, Simonsen, and Lawson 1995;Weerd and Day 2009). The pressure from citizens, 35 consumers and animal welfare organizations has been growing regarding animal rights, leading to changes in 36 the legislation. For example, the provision of manipulative materials to pigs of all ages is mandatory in the 37 European Union since January 2013 (Council Directive 2008/120/EC 2008), materials named as 38 'environmental enrichments'. Environmental enrichments are defined as materials susceptible to improve the 39 biological functioning of captive animals (Newberry 1995) and should stimulate species-typical animals' 40 sensory systems, cognitive capacities and behaviours (Wells 2009). For instance, for pigs, enrichments 41 materials should be edible, chewable, investigable, and manipulable (reviewed in Godyń, Nowicki, and Herbut 42 2019). Moreover, enrichment materials should be provided in such a way that they are of sustainable attraction 43 for pigs, and should be accessible for oral manipulation, and provided in sufficient amount (Newberry 1995;44 Godyń, Nowicki, and Herbut 2019). Enrichments effects are generally tested using behavioural and 45 physiological paradigms (Nannoni et al. 2016) and are classified as optimal (if they meet all of the above-46 mentioned criteria), suboptimal (if meet most of the criteria and should be combined with others) or marginal 47 (they do not fulfill the animal need and should be used with other) (Godyń, Nowicki, and Herbut, 2019). 48 In the particular case of pigs, abnormal patterns of behavior (stereotypies, belly nosing, ear and tail biting) 49 may arise at several stages of their development if they are prevented from any enrichment (Prunier et al. 50 2020). Enrichments have the potential to reduce these abnormal behaviours and increase positive behaviors 51 like play (Lykhach et al. 2020; Luo et al. 2020). Although straw bedding is one optimal enrichment according 52 to several literature references (reviewed in Godyń, Nowicki, and Herbut 2019), it is also non applicable for 53 many farms using slatted floor systems. Thus manipulative materials have been developed and are used in 54 farms in the form of ropes, hanging balls, wood, pipes or different commercial toys. 55 Besides those enrichment materials, one may wonder if enrichment may be provided by other stimuli in the 56 environment of farm animals. Pigs being social animals, social enrichment is sometimes used for lactating 57 piglets, by allowing different litters to interact. This enrichment enhances play and decreases aggression at 58 weaning (Salazar et al. 2018). Another relational partner of pigs is their caregiver. Human interactions seems 59 to correspond well to the definition of enrichment, i.e. they provide occasions of social contact with another 60 animal (stimulates biological functioning), and stimulate all sensory systems of the animals. Humans, notably 61 through their clothes and boots, are chewable, investigable, and manipulable. Many positive outcomes of 62 positive human interactions have been shown. Farm animals may be tamed by humans providing regular 63 positive additional contacts, leading to the expression of positive emotions (Tallet et al. 2018). Humans may 64 consequently be associated with positive outcomes as measured by a decrease of heart rate (Schmied et al. 65 2008; Tallet et al. 2014), modification of heart rate variability and ear postures (Coulon et al. 2015), EEG 66 (Rault et al. 2019). Humans can also acquire reassuring properties in situations of social isolation (Rushen et 67 al. 2001;Tallet, Veissier, and Boivin 2005). They may even induce behavioural reaction similar to those 68 toward social partners (Brajon, Laforest, Bergeron, et al. 2015). Regular positive human contacts may even 69 lead to improved welfare through positive cognitive bias (Brajon, Laforest, Schmitt, et al. 2015). In addition, 70 pigs raised in a poor environment may develop more interest toward a familiar human than pigs raised in an 71 enriched environment (Tallet et al. 2013), leading the author of the study to hypothesize a familiar human may 72 be perceived as an enrichment. To our knowledge, no comparison exists on the effect of object enrichment 73 and pseudo-social enrichment via human interactions. This would provide new insight into enrichment 74 practices in pig breeding, for improving their welfare. 75 In this study, we developed a paradigm to test the perception piglets may have toward two stimuli: an 76 inanimate object that could be used as enrichment, and a familiar human. The aim of this study was thus to 77 elucidate the specific value that a familiar human may have compared to an inanimate object of enrichment. 78 Twenty-four weaned piglets were previously similarly familiarized to both an object and a human. We first 79 evaluated the potential preference for one of the stimuli. Then we evaluated the emotional value of the stimuli 80 through isolation/reunion tests. We hypothesized that being reunited with a stimulus would attenuate the stress 81 of social isolation and promote positive behaviors, and even more that the stimulus has a positive emotional 82 value for piglets. We used behavioural but also bioacoustics data known to be relevant in comparing emotional 83 states of pigs (Friel et

Animals and experimental conditions 94
Twenty-four healthy weaned female piglets (Landrace/Large white dams inseminated with Piétrain semen) 95 were involved in total. Piglets were weaned at 28 days of age. Eight groups of three sister piglets from eight 96 different sows were selected at weaning according to their weight (the weight was balanced between and 97 within the groups, 9.05± 0.66 kg on average). Groups were reared in the same rearing room in 115 x 132 cm 98 pens, with slatted floor, isolated from each other by 1.5m high plastic panels. Piglets were fed ad libitum and 99 had continuous access to a water trough. Each pen was provided with a steel chain as enrichment. The piglets 100 were involved in the experiment from 28 to 57 days of age. 101 For different phases of the experiment, we also used an experimental room. The experimental room was 102 located in the same building as the rearing room, 15 m away, and was a 270 cm x 270 cm soundproof room. 103 It was warmed up by an electric heater. The entrance door was equipped with a hatch for the piglets. The 104 transportation from the rearing room to the experimental room was done by the usual caretakers in closed 105 trolleys. We used visually isolated trolleys to transport either the group of three piglets together (L120 x W80 106 x H80 cm) or one piglet at once (L80 x W50 x H80 cm). 107 3.3 Human and object familiarization 108 All the piglets were familiarized alternatively with two stimuli: an experimenter, referred as 'Human' in the 109 rest of the article (always the same, a 1.65 m high woman dressed with a blue overall) and a 5L-plastic can 110 (L20 x W10 x H30 cm) filled with water from which hung three pipe pieces tied with a thin rope, so that the 111 three piglets could chew it all together, referred as 'Object' in the rest of the article. Familiarizations started 112 at 28 days of age and ended at 53 days of age. Familiarizations were divided into two phases for each stimulus: 113 eight sessions in the home pen (from days 29 to35), and eight sessions in the experimental room (from days 114 36 to 43 and from days 49 to 53). In the home pen, piglets received 10 minutes sessions twice a day for each 115 stimulus along four days. During the same week, all groups were alternatively transported to the empty 116 experimental room and remained there for 10 minutes, once a day for 5 days to get familiarized. After this 117 familiarization, piglets received 10 minute sessions in the experimental room once a day for each stimulus, 118 along nine days. The same procedure was used for each group of three pen mates, as follow: 119 •   Object: the experimenter came before the grid of the pen holding the object and stood still and quiet  120  for 30 seconds. Then, she entered the pen to tie the object to the grid opposite to the entrance with a small  121 rope and went out. From the moment she went out, the object was left for 10 min in the pen. Then the 122 experimenter removed the object. 123 • Human: the experimenter came in front of the grid of the pen and stood still, holding a stool, and quiet 124 for 30 seconds. She then entered to sit on the stool in the pen, close to the grid at the opposite of the entrance. 125 During the first session of human familiarization, the experimenter sat still without moving for 10 minutes 126 before going out, in order not to afraid the piglets. For the other familiarization sessions, she engaged in 127 interactions with each piglet (similar protocol as in Tallet  The piglet is considered in a virtual zone when its forelegs and head are in the zone. Number of changes (2) 1

Located in stimulus zone
The animal is considered in a zone when its forelegs and head are in the zone.

Located in proximal zone
The piglet is considered in the proximal zone when its forelegs and head are in the zone.
Total time (2) 8 Latency to first entrance (2) 9 Located in distal zone The piglet is considered in the distal zone when its forelegs and head are in the zone.
Total time (2) 10 Postures At 55, 56 and 57 days of age, piglets were confronted to an Isolation/Reunion test in order to assess their 172 perception of the stimuli and its potential to calm the piglets after a period of stressful isolation (Fig. 1B). The 173 test consisted in two phases. The piglet was brought individually in a trolley to the experimental room, the 174 hatches were opened and the piglet was gently pushed into the room. It was left alone for five min, which 175 defined the 'Isolation' phase. Then, one of the stimulus ('Human', 'Object') or 'Nothing' was shown to the 176 piglet for 30 seconds: the door was opened with either: a) the human standing with the stool, b) the human 177 standing with the object, or c) nothing. Finally, the second phase named 'Reunion' phase occurred and 178 consisted in either a) presence of the experimenter sitting in the room on a stool and remaining immobile and 179 quiet, b) presence of the object tied in the room, or c) isolation in the room for 270 seconds. Each piglet was 180 thus tested three times, with one test per piglet per day. The order of the modalities (reunion with human, 181 object or nothing) was randomized within the days and between the piglets of a same pen. 182 3.5.2 Behavioural analyses 183 The tests were video recorded (camera Bosh, Box 960H-CDD) and a recorder (Noldus Mpeg recorder 2.1., 184 The Netherlands), linked to a LCD monitor (DELL 48 model 1907FPc) which allowed us to visualize the test 185 from an adjacent room. The location of the piglets was monitored directly during the tests and the other 186 behaviors later from videos, both using the software The Observer XT 14.0 (Noldus®, The Netherlands). All 187 behaviors used are indicated in table 1 and correspond to numbers : 1, 8-10, 14. 188 3.5.3 Acoustic monitoring and analyses 189 Vocalizations produced during the Isolation/Reunion test were recorded with a C314 microphone (AKG,190 Austria) placed in the center of the room and at one-meter-high, connected to a MD661MK2 recorder 191 (Marantz, USA). The vocal types were then manually annotated (grunt, squeak, bark, scream and mixed calls), 192 after visual inspection of spectrograms on Praat® software, by an expert. Only grunts were subsequently 193 acoustically analyzed as they represented the most frequent call type that constituted a dataset of 5766 calls. 194 A spectro-temporal analysis was performed with custom-written codes using the Seewave R package (Sueur,195 Aubin, and Simonis 2008) implemented in R (R Core Team 2015). After a 0. To test whether the first approach depended on a stimulus, it was tested in a binomial model (Human or  225 Object) and the effect of the day and the position of the human were put in an additive model (model 2). The 226 number of times a piglet laid down in the proximal zone close of a stimulus was tested as a binomial variable 227 (presence vs. absence) and a 2 test was used. 228 3.6.2 Analysis of Isolation/Reunion tests: spatial and vocal behavior of piglets 229 3.6.2.1 Behavioural response scores 230 To be able to have comparable behaviors, between phases and stimuli, and to reduce the number of variables, 231 behaviors were gathered and a PCA was computed (parameters: 1, 8-10, 14, To be able to compare the spectro-temporal structure of grunts, two scores were built. The duration of grunts 240 was log transformed and used as a temporal score (linear distribution). For spectral analysis, parameters 241 previously extracted were gathered in a PCA to be able to monitor which parameters load the same way and 242 build an acoustic score. Only one PC had an Eigen value above one, explained 83% of the variability and was 243 named 'Acoustic spectral score (PCac.)'. The three behavioural response scores (IsoReuPC1, IsoReuPC2, IsoReuPC3) and the two acoustic scores 247 (PCac. and log(duration)) were used as response variables in a linear model testing i) the two-way interaction 248 between the type of reunion (Human/Object/Nothing) and the phase of the test (Isolation/Reunion), ii) the 249 two-way interaction between the day of the test (1/2/3) and the phase iii) the day of the test and the type of 250 reunion (model 3). 251 3.6.3 Analyses of predictors for vocal expression during the reunion with the stimulus 252 The aim of this analysis was to search for the best predictors of vocal dynamic and grunt acoustic features, in 253 presence of the human or the object. For this analysis, only the dataset containing the reunion phases with the 254 Human or the Object were used, extracted from the Isolation/Reunion test. 255 3.6.3.1 Monitoring spatial proximity toward the stimulus and time during the test 256 The location of the piglet in the room was divided into two categories: when the piglet was located in the 257 proximal zone ('Close') and when the piglet was located elsewhere ('Away') to build a two level factor named 258 'Location'. This parameter allowed us to track for spatial proximity toward the stimulus. continuous variable), the spatial proximity of the piglet toward the stimulus (two levels: close to or away from 277 the stimulus), the behavioural proximity of the piglet toward the stimulus (continuous behavioural proximity 278 score) or interactions between the type of stimulus and the location, between the type of stimulus and the 279 behavioural proximity toward the stimulus or between the type of stimulus and the time along the test. To 280 search of the best predictors of vocal expression, five acoustic variables were used as response variables. 281 Three variables were linear: the acoustic spectral score PCac., the duration of grunt (log(grunt duration)) and 282 the grunt rate (number of grunt per second, calculated when the number of grunts per interval was above three 283 (186 intervals out of 286 interval). Two non-linear variables were used: the total number of grunts (Poisson 284 distribution), the number of times grunts were produced in series (Binomial distribution), i.e when more than 285 one grunt was produced in a given interval. Indeed, since we used only interval containing at least three grunts 286 to calculate the grunt rate, we needed to control we were not missing information on interval containing fewer 287 grunts, so we used the occurrence of one grunt interval to counteract the effect of interval selection. 288 A full model, describing the experimental design, was built as follow ('lmer' or 'glmer' function of 'lme4' R 289 package): Model 4 = Response variable ~ day + stimulus + location + Z interval index + Z behavioural 290 proximity score + stimulus*location + stimulus*Z behavioural proximity score + stimulus*Z interval index 291 + location*Z behavioural proximity score +(1|individual). To increase interpretability, all continuous 292 variables (interval index and behavioural proximity scores) were scaled, so the Z score is presented every 293 time (Schielzeth 2010), the individual was put as random factor to take into account multiple tests of the same 294 piglet. On this full model, a model selection was performed with the 'dredge' function of the 'MuMIn' R 295 package (Bartoń 2016), which compares all possible models built using subsets of the initial explanatory 296 variables of the full model, including null model. Models were compared using Akaike Information Criteria 297 corrected for small sample size (AICc). Significant models were selected when delta AICc was below two 298 (Burnham and Anderson 2002), the weight of remaining explanatory variables was evaluated by calculating 299 the presence or absence of the term in the remaining models ('importance' function). It has to be noticed that 300 for the occurrence of one grunt intervals (Binomial model) no significant models were selected since the null 301 model was contained in the best selected models (AICc<2  The first behavioural response score (choicePC1, 45.8%) was negatively correlated with the mean duration in 324 stimulus zone, proportion of time spent in stimulus zone, time spent in contact with the stimulus and the total 325 time spent in zone. Statistics revealed a significant effect of the interaction between the type of stimulus and 326 the day of the test ( 21 = 6.3, p = 0.012), but post hoc tests did not show any difference between groups 327 (pairwise tests with Tukey correction, t.ratio <|2.2|, p>0.15, fig. 2A). The second behavioural response score 328 (choicePC2, 19.5%) was positively correlated with the number of visit in stimulus zone. Statistics showed no 329 interaction between the type of stimulus and the day of the test ( 21 = 0.7, p = 0.4), a trend for the effect of the 330 day ( 21 = 3.3, p = 0.07) and a main effect the type of stimulus: PC2 was higher when considering the human 331 zone compared to the object zone ( 21 = 7.3, p = 0.007, fig. 2B). The third behavioural response score 332 (choicePC3, 16%) was negatively correlated with the time the piglet spent exploring the stimulus zone and 333 the latency to approach the stimulus zone. Statistics showed no effect of explanatory variables on choicePC3 334 (Stimulus: 21 = 1.5, p = 0.2, Day: 21 = 0.6, p = 0.5). We counted the number of times the human zone or the 335 object zone was first approached by the piglet during the test. Statistic showed a trends for the effect of the 336 day of the test: piglets tended to approach first the object zone more often the second day than the first day of 337 the test ( 21 = 3.4, p = 0.06, fig. 2C). Last, we counted the number of times piglets laid down near a stimulus 338 zone, a 2 showed a significantly different distribution of occurrences of this behavior, that only occurred in 339 the human zone (by nine individuals out of 24) and none in the object zone ( 2 1= 12.8, p < 0.001, fig. 2D). 340 The position of the human in the room (left or right side) was included in all models and never showed a 341 significant effect (see supplementary The first behavioural response score (IsoReuPC1, 31.5%) was negatively correlated with the time spent 357 immobile, the time spent in distal zone and the number of changes of virtual zone. Statistics revealed a 358 significant interaction between the type of reunion and the phase of the test ( 22 = 16.6, p < 0.001, fig. 3A). 359 During the isolation phase, no significant difference was found between groups (pairwise comparisons 360 human/object/nothing, |t.ratio| < 0.7, p > 0.9) whereas during the reunion phase, the three type of reunions 361 differed significantly in PC1 values (human vs. object: t.ratio = 3.1, p = 0.03, human vs. nothing: t.ratio = 6.3, 362 p < 0.001, object vs. nothing: t.ratio = 3.3, p = 0.02). The reaction to each reunion type did not have the same 363 amplitude too. When piglets were not reunited with a stimulus, statistics did not show differences between the 364 isolation and the reunion phases (isolation vs reunion with nothing: t.ratio = 0.6, p > 0.9), whereas when 365 reunited with a stimulus, IsoReuPC1 showed a significant increase that was stronger with the human (isolation 366 vs. reunion, t.ratio = -6.3, p < 0.001) than with the object (isolation vs. reunion, t.ratio = -3.2, p < 0.03). 367 The second behavioural response score (IsoReuPC2, 39%) was positively correlated with the time spent 368 exploring the room and negatively correlated with the time spent looking at the entrance door and the latency 369 to enter the proximal zone of the stimulus. A significant interaction was found between the type of reunion 370 and the phase of the test ( 22 = 41.5, p < 0.0001, fig. 3B). During the isolation phase, no significant difference 371 was found between groups (pairwise comparison human/object/nothing, |t.ratio| < 2.7, p > 0.08) whereas 372 during the reunion phase, the two types of stimuli differed significantly (human vs. object: t.ratio = 4.9, p < 373 0.001), as well as the reunion with the human compared to nothing (t.ratio = 6.8, p < 0.001), but no difference 374 was found when comparing reunions with the object or nothing (t.ratio = 2.0, p = 0.37). The reaction to the 375 three types of reunions differed too: from isolation to reunion phase, no difference was found in IsoReuPC2 376 when piglets were reunited with the human (t.ratio = -0.6, p = 0.9), whereas PC2 decreased significantly when 377 piglets were reunited with the object or with nothing (object: t.ratio = 3.8, p = 0.003, nothing: t.ratio = 8.5, p 378 < 0.001). 379 The third behavioural response score (IsoReuPC3, 11.5%) was negatively correlated with the time spent in 380 the stimulus zone. Statistics showed a significant interaction between the type of reunion and the phase of the 381 test on IsoReuPC3 ( 22 = 36.4, p < 0.001, fig. 3C). During the isolation phase, no significant difference was 382 found between groups (pairwise comparison human/object/nothing, |t.ratio|<0.7, p>0.9). During the reunion 383 phase, for piglets being reunited with nothing, IsoReuPC3 differed significantly compared to being reunited 384 with a stimulus (human vs. nothing: t.ratio = -5.7, p < 0.001, object vs. nothing: t.ratio = -7.8, p < 0.001) but 385 IsoReuPC3 between the two types of stimuli did not differ (human vs. object: t.ratio = 2.1, p = 0.3). The 386 reaction to the three types of reunions differed too: from isolation to reunion phase, no difference was found 387 in IsoReuPC3 when piglets were not reunited with a stimulus (reunion with nothing: t.ratio = -0.8, p = 0.9), 388 whereas IsoReuPC3 decreased significantly when piglets were reunited with the object or with the human 389 (object: t.ratio = 7.6, p < 0.001, human: t.ratio = 4.8, p < 0.001). 390 The day of the test did not show any effect on IsoReuPC2 and IsoReuPC3 ( 22 = 0.9, p = 0.6, 22 = 0.2, p = 391 0.9 respectively) but was significantly higher for IsoReuPC1 from day one to day three ( 22 = 10.1, p = 0.007, 392 supplementary figure S1A). Post hoc test showed the differences in IsoReuPC1 were progressive along days 393 (pairwise comparison, day 1 vs. day2: t.ratio = -2.4, p = 0.05, day 1 vs. day 3: t.ratio =-3.0, p = 0.009, day 2 394 vs. day3: t.ratio = -0.6, p = 0.79). 395  All 5766 grunts produced during the test were analyzed using two acoustic scores: the logarithm of grunt 405 duration and a spectral score, computed from the PCA of frequency and noise parameters of the calls (acoustic 406 spectral score PCac., variable loading table 4): the greater the score, the lower the frequency and the lower the 407 spectral noise in the grunt. 408 Acoustic spectral score (PCac.)  fig. 3D). During the isolation phase, no difference was found 411 between groups (pairwise comparison during isolation, human/object/nothing: |t.ratio| < 1.9, p > 0.4) whereas 412 during the reunion phase significant differences were found between groups (pairwise comparisons during 413 reunion, human vs. object: t.ratio = -4.9, p < 0.001, human vs. nothing: t.ratio = -9.2, p < 0.001, nothing vs. 414 object: t.ratio = 3.7, p = 0.003). The reaction to each of the reunion types did not have the same amplitude too. 415 When piglets were subjected to another isolation, statistics did not show differences between the isolation and 416 the reunion phase (t.ratio = 0.03, p = 1), whereas when reunited with a stimulus, PCac. showed a significant 417 decreased that was stronger with the human (t.ratio = 9.3, p < 0.001) than with the object (t.ratio = 5.3, p < 418 0.001). Statistics also showed a significant interaction between the type of reunion and the day of the test ( 21 419 = 26.8, p < 0.001) but post hoc tests revealed no significant pairwise comparisons (|t.ratio| < 1.6, p > 0.8, see 420 supplementary table S4-S6 and supplementary fig. S1B). 421 Grunt duration showed a significant interaction between the type of reunion and the phase of the test ( 22 =  422 210.1, p < 0.001, fig. 3E). During the isolation phase, no difference was found between groups (pairwise 423 comparison during isolation, human/object/nothing: |t.ratio| < 2.6, p > 0.09) whereas during the reunion phase 424 significant differences were found between groups (pairwise comparisons during reunion, human vs. object: 425 t.ratio = -19.5, p < 0.001, human vs. nothing: t.ratio = -16.7, p < 0.001, nothing vs. object: t.ratio = -3.9, p = 426 0.003). The reaction to each of the reunion types differed too. When piglets were subjected to another isolation 427 or reunited with the object, statistics did not show differences between the isolation and the reunion phase 428 (pairwise comparisons isolation vs. reunion object/nothing: |t.ratio| < 0.6, p > 0.6), whereas when reunited 429 with the human, grunt duration decreased significantly (pairwise comparison isolation vs. reunion, human: 430 t.ratio = 9.3, p < 0.001). Last, statistics also revealed a significant main effect of the day of the test ( 22 = 20.0, 431 p < 0.001): grunt duration decreased as the day of the test increased, especially between the first two days 432 (pairwise comparisons, day 1 vs. day2: t.ratio = 3.9, p < 0.001, day 1 vs. day 3: t.ratio = 2.6, p = 0.03, day 2 433 vs. day 3: t.ratio = -1.2, p = 0.4, supplementary figure fig. S1C). 434

Effect of proximity to stimulus on vocal expression 435
The four following acoustic variables: total number of grunts, grunt rate, duration of grunts (log(grunt 436 duration)) and spectral acoustic score (PCac.) may be predicted by the context (the type of stimulus), the 437 spatial proximity to the stimulus (location in the room), variables independent from the stimuli (day, time 438 along the test, described by the interval index) or the experience piglets previously had with the stimuli. To 439 quantify the experience piglets had with each stimulus (closeness and exploring), a behavioural proximity 440 score resulting two from principal component analyses were built ( table 5) and one was selected per type of 441 reunion: 'behavioural proximity score' corresponded to the opposite of HproxPC1/OproxPC1 (respectively 442 for reunion with the human or the object) and was positively correlated with the time spent in contact with 443 and near the stimulus. After model comparison and selection of the best equivalent models, the weight of 444 predictors as well as the estimates of the averaged resulting model were calculated (tables 6 and 7, 445 respectively, full selected models in supplementary

458
The model selection showed the total number of grunts was predicted by the interactions between the type of 459 stimulus and the location of the piglet in the room, as well as the interaction between the type of stimulus and 460 the behavioural proximity score (table 6). Thus, a lower number of grunts was likely to predict the piglet was 461 reunited with the object, and spatially close to it ( fig. 4B). In addition, when reunited with the object, the 462 higher the behavioural proximity score (-OproxPC1), the higher the probability of producing more grunts ( fig.  463 4A), but not with the human. Concerning grunt rate, the type of stimulus was the only consistent predictor 464 (table 6): when piglets were reunited with the human, the rate of grunt was higher than when reunited with 465 the object (fig. 4C). 466 Considering the acoustic structure of grunts (duration and spectral acoustic score PCac.), both descriptors 467 were best predicted by the interaction between the location in the room and the type of stimulus, the 468 behavioural proximity score, the interval index and the day (table 6). The probability of having shorter grunts 469 was higher when reunited with the human and close to her ( fig. 4D). In addition, the higher the behavioural 470 proximity score, the higher the probability of having shorter grunts ( fig. 4E). The probability of having longer 471 grunts increased as the time of the test increased (interval index, fig. 4F) with no interaction with the type of 472 stimulus or location. Last, as the day of the test increased, the probability of having shorter grunt increased 473 (slope estimate ± se: -0.13 ± 0.04 and -0.04 ± 0.04, respectively for day two and three, of producing grunts with a lower acoustic spectral score depended of the type of stimulus and the spatial 476 proximity, as the acoustic spectral score was more likely to decrease when approaching the object but not the 477 human ( fig. 4G).The higher the behavioural proximity score, the higher the probability of producing grunt 478 with a lower acoustic spectral score, independently from the type of stimulus and location in the room ( fig.  479 4H, table 6). As the time during the test increased, the probability of producing grunts with a higher acoustic 480 spectral score increased, independently from the type of stimulus or location ( fig 4I). Last, as the day of the 481 test increased, the probability of producing grunts with a lower acoustic spectral score increased independently 482 for the type of stimulus (slope estimate ± se: -067 ± 0.26 and -0.51 ± 0.27, respectively for day two and 483 three,    In a V shaped apparatus Choice test, comparing the time spent close to and in contact with each of the stimulus 488 (first behavioural response score, choicePC1) or the latency to reach the stimulus zone and exploring the 489 stimulus zone (third behavioural response score, choicePC3), did not lead to significant differences between 490 the types of stimuli. No evidence for a preference in the first approach could be found too. So this test did not 491 allow to conclude for a preference for one of the stimuli. The novel conformation of the testing pen compared 492 to what pigs used to know (open pen) may have attracted their attention more than the familiar stimuli present. 493 Using the home pen to test for preference, like done in mice for instance (Blom et al. 1992 our study, two differences came out between the human and the object. The piglets entered more often the 501 human zone than the object zone (second behavioural response score, choicePC2) and the number of times 502 piglets laid down near the stimulus was human zone specific. Laying down is a sign of absence of stress in 503 pigs (Goumon and Špinka 2016), and the location was not by chance here. We may hypothesise that the human 504 had reassuring effects, as it has been found in studies on other farm animal species (Rushen et al. 2001;Tallet, 505 Veissier, and Boivin 2005). This will need to be confirmed by monitoring hear rate and its variability. 506 5.1.2 Behavioural evidence for positive attractivity of both a manipulable object and the familiar human 507 Isolation/reunion tests allowed us to show differential responses according to the stimulus piglets were 508 reunited with. Behavioural measures showed that stimuli were attractive for the piglets with a decrease in the 509 time spent in the zone distant from the stimuli and an increase in the time spent in the stimulus zone when 510 piglets were reunited with the human or the object compared to remaining alone in the experimental room. 511 Also piglets remained less immobile inactive, a behavior associated with the vigilance response, and they had 512 a lower locomotor activity when one stimulus was present. These changes in locomotor activity have to be 513 explained along with the time spent near the stimulus and are in line with the hypothesis of attractivity to the 514 stimuli. Beyond this general changes in behaviour, piglets expressed discriminative behaviours according to 515 the stimulus present. Indeed, in response to a reunion with the human compared to the object, piglets were 516 quicker to enter the stimulus zone, expressed a lower mobility and a higher exploration time. In response to a 517 reunion with the object, piglets spent more time watching the exit door than exploring the room, a response 518 equivalent to the reunion with nothing (i.e. isolation). As a consequence, results may show the presence of the 519 familiar human may prevent the piglets from expressing stress responses (more vigilance behaviour and less 520 exploration), a hypothesis strengthened by the fact that being reunited with the object or nothing seems 521 equivalent in terms of vigilance behviors. 522 5.1.3 Acoustic evidence of a high arousal positive emotional state with the human and a low arousal emotional 523 state with the object 524 We predicted that if vocalisations allow expression of the emotional state of the piglets, acoustic scores should 525 be different when piglets would be reunited with a stimulus compared to nothing. In reaction to the reunion 526 with the familiar human, the duration of grunts decreased and this was not the case with the object or when 527 piglets remained alone. Shorter vocalisations have been associated with positive contexts compared to 528 negative ones in many species (Briefer 2012), and specially shorter grunts in pigs (Friel et  comparisons has to be taken precociously due to material and methodological specificity of each study, the 534 range of values we obtained with the human here are in the range of positively perceived situations. Behavior 535 and acoustic together may allow to conclude that being reunited with the human leads to a more positive 536 context than with the object. Since the human has been associated with positive tactile contacts, known to 537 promote a positive state (Tallet et al. 2014;, approaching the human may engage piglets in a positive 538 anticipation of positive tactile interactions. The object rather leads to an emotional state having a valence 539 comparable to isolation, even if it is attractive to some extent. During reunion with either the object or the 540 human, the spectral acoustic score of piglets grunts decreased: grunts were composed of higher frequencies 541 and a higher noise component and this effect was greater with the human compared to the object. Changes in 542 spectral components in response to changing contexts may be associated with the arousal of situations in 543 mammals (Ehret 2018;). This may underline that the reunion with a stimulus promotes 544 emotional states of high intensity in piglets, especially the reunion with a human. Villain et al. (2020) showed 545 piglets were able to rapidly change the spectral properties of their grunts when anticipating positive events. 546 The anticipation of a reunion with familiar conspecifics led to noisier grunts whereas the anticipation of a 547 reunion with a familiar human associated with positive contacts, led to higher pitched grunts. In the present 548 study, frequency and noise components of the grunt are closely intercorrelated, so it is not possible to 549 discriminate between the two. To summarize, for isolated piglets, being reunited with a familiar human 550 induces a high arousal positive emotional state, while being reunited with a familiar object induces a low 551 arousal emotional state. Thus, human positive interaction seems to be more valuable as an enrichment for 552 piglets. This may result from the relationship created through the numerous sessions of positive vocal and 553 tactile interactions, as already shown in previous studies (Tallet et al. 2014;Brajon, Laforest, Schmitt, et al. 554 2015). An inanimate object may not acquire similar properties. As a consequence promoting social or pseudo-555 social enrichment in pigs is a good way to enhance their welfare. 556 5.1.4 Experience and spatial location predict differences in spectro-temporal features of grunts depending on 557 the stimulus 558 To go further we studied which variables predicted vocal production. From the model selection, we found that 559 the type of stimulus (object or human) was among the best predictors of vocal expression (number of grunts, 560 grunt rate, duration, and spectral score) and was the only consistent predictor explaining the temporal dynamic 561 (grunt rate). Being reunited with a human (and not an object) is associated with more vocal production and at 562 a higher rate. Morton (1977) explained that the rhythm of a behavior can be positively linked to motivation of 563 the producer, and thus a higher arousal.  showed piglets had a higher grunt rate when 564 anticipating the arrival of conspecifics, compared to a familiar human. Here, we would interpret the result in 565 the direction of a higher motivation toward the human compared to the object. 566 Being reunited with the human and close to them is likely to induce shorter grunts whereas being reunited 567 with the object and close to it is likely to induce a fewer number of higher frequency and noisier grunts. This 568 is in line with the positive state induced by the human compared to the object. 569 The behavioural proximity, behavioural score associated with the number of interactions and the time spent 570 in contact with or near the stimulus was a consistent predictor for both acoustic scores. Independently from 571 the type of stimulus, the higher the behavioural proximity to the stimulus was, the higher the probability for 572 producing shorter grunts with higher frequencies and noise components was. This raises the question of the 573 possibility to monitor the degree of behavioural proximity to an enrichment using the structure of grunts 574 (Briefer et al. 2019). 575 The time during the test was also a predictor of the spectro temporal features: the later in the test, the higher 576 the probability of producing longer, lower pitched and less noisy grunts was (effect of interval index). Positive 577 effect of stimulus presence may be attenuated with time, and/or negative effects of isolation from penmates 578 may increase. This was independent from the type of stimulus, we can hypothesize that during the test, since 579 the human do not interact with the piglet (as it would outside the test) make the human closer to an object and 580 piglets may habituate to the stimulus. It would be interesting to investigate whether interacting with the piglet 581 during an Isolation/Reunion test may prolong the positive effect of the reunion with a familiar human after a 582 five minute isolation. Last, along days, grunts were more likely to be shorter, higher pitched and noisier, 583 independently from the type of stimulus. This may have to be linked to habituation to the protocol along test. 584 5.1.5 A familiar human as enrichment: implication for pig welfare 585 Reunion with a familiar object or human led to an attraction toward the stimulus and repeated contacts, as 586 well as a decrease in vigilance behavior. These parameters are in line with the definition of what a 587 'environmental enrichment' should promote (Godyń, Nowicki, and Herbut 2019; Newberry 1995). Both 588 stimuli we provided can be considered as enrichments. Being with a human and/or close to the human 589 provokes higher degrees of behavioural changes in piglets (both spatial and vocal) and specific behavioural 590 postures (laying), associated with positive states. Regarding vocal behavior, although we showed that the 591 behavioural proximity to the stimulus and vocal responses correlated, only the human presence led to positive 592 shorter grunts during the reunion, and not the object. Regarding enrichments, it has been pointed out that the 593 novelty is a paramount feature to promote a long term positive context and delay habituation effects (Trickett,594 Guy, and Edwards 2009). It is possible that interactions with the human may allow this feature, indeed, a 595 human is moving, talking and is not likely to reproduce exactly the same gesture, at the same rhythm, which 596 may participate in keeping a higher level of stimulation than an object may do. More studies are needed to 597 better describe what are the human most efficient signals and behaviors that promote positive emotional states 598 using a multimodal approach: To conclude, using behavioural and vocal monitoring, this study showed that a manipulable object and a 603 familiar human can be attractive for weaned piglets out of their rearing environment. A familiar human may 604 enhance a positive and high arousal emotional state when the piglet is alone, mimicking reassuring effects. 605 Humans may be considered as enrichment in the piglets' environment and more studies should consider 606 pseudo social interactions between humans and piglets to enhance welfare. In order to be applicable on a 607 larger scale, the kinetic of human-piglet relationships needs to be better understood, as well as the most 608 efficient signals triggering positive emotional states in piglets.