Zoo housed meerkats do not recognise human emotions (Supporting data)

Supporting dataset and code for [citation when paper is published]. Abstract Domestic mammals can categorically process and respond functionally to human emotional cues. However, to date, it is unclear whether this ability is underlaid by shared mammal emotional processing, associative learning through exposure to humans, or has arisen through the domestication process. To disambiguate these hypotheses, we investigated whether zoo-housed meerkats (Suricata suricatta), who have not undergone the process of domestication but are regularly exposed to human contact, recognise human emotion, and whether this is modulated by human interaction levels. 35 meerkats’ were presented with happy, sad, fearful and neutral human emotional displays to assess their functional responses. A cross-modal paradigm also tested for expectancy violation responses to emotionally incongruent signal pairs. While isolated findings suggest human interaction level modulated some behavioural responses to the test paradigms, no patterns of functional responses, lateralised processing, or cross-modal integration consistent with those documented in domestic mammals emerged, either for the whole study population or for those meerkats with a history of close interaction with humans. Thus, zoo-housed meerkats do not show strong evidence of human emotion recognition. Our findings support the hypothesis that the capacity for human emotion recognition in domestic species has arisen through artificial selection for attention to human communicative cues. Supporting data The dataset used to carry out the analysis reported in this paper is available in HEAT_RQ1_Final_Analysis_Dataset.xlsx. Variables included Human emotional signal presentation trial data Obs_ID: Unique identifier for each observation Media: Stimulus medium: audio only ("Audio"), video only ("Video") or audiovisual ("Cross_Modal") Zoo: Zoo identifier Meerkat: Meerkat identifier Actor: Actor identifier - numbering matches the identifiers in the RAVDESS database Session: Session number, within each stimulus medium block. Audio only and video only blocks comprised 3 sessions each, while the cross modal block comprised 6 sessions Trial: Trial number within the session, for each individual Emotion: Emotion being portrayed by the actor: takes values "happy", "angry", "fearful" and "neutral". For cross-modal stimuli, Emotion corresponds to the emotion portrayed in the *video* (with matching audio if Congruence = "Match", or non-matching audio if Congruence = "Mismatch") Congruence: For cross-modal stimuli, indicates whether audio and visual emotional stimuli match ("Match) or don't ("Mismatch"). Recorded as NA for audio only and video only blocks Valence: Valence of the emotional signal - Happy and Neutral emotions were considered to have Positive valence, while Angry and Fearful emotions were considered to have Negative valence Zoo-level data: HAI_Level: Human-animal interaction level, for each zoo. Zoos where meerkats did not take part in encounters were coded as 1, corresponding to lower levels of HAI, while zoos where meerkats took part in encounters were coded as 3, corresponding to higher levels of HAI. Individual vigilance data based on behavioural observations of spontaneous behaviour Vigilance_Perc: Percentage of time spent vigilant, averaged across observations for each individual Human_Vigilance_Perc: Percentage of time when vigilance was directed towards humans out of the total time being vigilant, averaged across observations for each individual Behavioural responses to human emotional signal presentation Tube: Time spent in the tube (seconds) Zone_3: Time spent in the zone of the test arena furthest from the laptop (zone 3) (seconds) Zone_2: Time spent in the middle zone of the test arena (zone 2) (seconds) Zone_1: Time spent in the zone of the test arena closest to the laptop (zone 1) (seconds) Binocular_Look: Time spent with both eyes visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the field of binocular vision Left_Gaze: Time spent with the left eye visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the left field of vision Right_Gaze: Time spent with the right eye visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the right field of vision No_Look: Time spent with no eyes visible on the footage from the GoPro directly above the laptop screen, ie with the screen not in the field of vision In: Total time spent in the test arena (zone 1, 2 and 3) Lat_Index: Laterality index, defined as LI= (Time spent on left gaze-Time spent on right gaze)/(Total looking time (left+right+binocular)) Binocular_Look_Perc: Percentage of time in the test arena with both eyes visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the field of binocular vision Left_Gaze_Perc: Percentage of time spent in the test arena with the left eye visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the left field of vision Right_Gaze_Perc: Percentage of time spent in the test arena with the right eye visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the right field of vision No_Look_Perc: Percentage of time spent in the test arena with no eyes visible on the footage from the GoPro directly above the laptop screen, ie with the screen not in the field of vision Look_Perc: Percentage of time in the test arena withat least one eye visible on the footage from the GoPro directly above the laptop screen, ie with the screen in the field of vision Z1_Pres: Presence or absence in the zone closest to the screen, coded as Yes or No Supporting code In addition to the supporting dataset, we supply: Analysis code (R): the R code used to carry out the analysis reported in the paper and create Figures 3, 4 and 5. Available in HEAT_RQ1_Final_Publication_Analysis_Code.R Trial Runner code (Python): the Python/PsychoPy based code used to run emotion recognition testing sessions. For each trial, this code accessed data collection records for the relevant subject, chooses an emotion at random out of emotions not yet seen during the session, displays the emotional stimulus (audio, visual or audio-visual media), and adds a new row to data collection records to record subject ID and group, the nature of the stimulus presented, and date and time of presentation. Audio-only sessions: code available in reworked_audio_runner.py Visual-only sessions: code available in reworked_video_runner.py Cross-modal sessions: code available in reworked_cross_modal_runner.py本数据集为论文发表后将补充引用标注的配套数据集。 摘要 家养哺乳动物可对人类情绪线索进行分类化处理并做出功能性响应。然而迄今为止，学界仍不清楚该能力的底层机制究竟是共享的哺乳动物情绪加工通路、通过接触人类形成的联想学习，还是驯化过程所催生的结果。为厘清上述三种假说，本研究针对未经历驯化但定期接触人类的圈养狐獴（*Suricata suricatta*）展开实验，探究其是否具备人类情绪识别能力，以及该能力是否受人类互动水平的调控。研究向35只狐獴展示人类表现出的快乐、悲伤、恐惧与中性情绪表情，以评估其功能性响应；同时采用跨模态范式，测试其对情绪不一致信号对的预期违背响应。尽管个别研究结果表明，人类互动水平会对部分测试范式下的行为响应产生调控作用，但无论是针对整个研究种群，还是针对与人类存在密切互动历史的狐獴个体，均未出现与家养哺乳动物文献记载相符的功能性响应、偏侧化加工或跨模态整合模式。因此，圈养狐獴并未表现出人类情绪识别能力的有力证据。本研究结果支持如下假说：家养物种的人类情绪识别能力，是通过对人类交际线索注意力的人工选择演化而来的。 变量说明 人类情绪信号呈现试验数据 Obs_ID：每条观测记录的唯一标识符 Experiment：刺激呈现媒介，可选仅音频（"Audio"）、仅视频（"Video"）或视听跨模态（"Cross_Modal"） Zoo：动物园标识符 Meerkat：狐獴个体标识符 Actor：演员标识符，其编号与RAVDESS数据库（Ryerson Audio-Visual Database of Emotional Speech and Song）中的标识符一致 Session：每个刺激媒介区块内的实验会话编号。仅音频与仅视频区块各包含3个会话，跨模态区块则包含6个会话 Trial：每个个体在单一会话内的试次编号 Emotion：演员所表现的情绪类型，取值为"happy"（快乐）、"angry"（愤怒）、"fearful"（恐惧）与"neutral"（中性）。对于跨模态刺激，Emotion对应视频中呈现的情绪（若一致性（Congruence）为"Match"（匹配），则音频情绪与之匹配；若为"Mismatch"（不匹配），则音频情绪与之不一致） Congruence：针对跨模态刺激，指示视听情绪刺激是否匹配（"Match"为匹配，"Mismatch"为不匹配）。仅音频与仅视频区块的该字段记为NA Valence：情绪信号的效价——快乐与中性情绪被归为正性效价，愤怒与恐惧情绪被归为负性效价 动物园层级数据 HAI_Level：各动物园的人兽互动（Human-Animal Interaction, HAI）水平。狐獴不参与游客互动的动物园编码为1，对应较低的HAI水平；狐獴参与游客互动的动物园编码为3，对应较高的HAI水平 基于自发行为观测的个体警戒行为数据 Vigilance_Perc：个体警戒时长占总观测时长的百分比，取该个体所有观测记录的平均值 Human_Vigilance_Perc：个体警戒行为中朝向人类的时长占总警戒时长的百分比，取该个体所有观测记录的平均值 人类情绪信号呈现下的行为响应 Tube：在通道内停留的时长（秒） Zone_3：在测试场地远离笔记本电脑的区域（区域3）停留的时长（秒） Zone_2：在测试场地中部区域（区域2）停留的时长（秒） Zone_1：在测试场地靠近笔记本电脑的区域（区域1）停留的时长（秒） Binocular_Look：笔记本电脑屏幕正上方的GoPro拍摄画面中，狐獴双眼均可见的时长，即屏幕处于双眼视野范围内的时长 Left_Gaze：GoPro拍摄画面中，狐獴左眼可见的时长，即屏幕处于左眼视野范围内的时长 Right_Gaze：GoPro拍摄画面中，狐獴右眼可见的时长，即屏幕处于右眼视野范围内的时长 No_Look：GoPro拍摄画面中，狐獴双眼均不可见的时长，即屏幕未处于其视野范围内的时长 In：在测试场地（区域1、2、3）内的总停留时长 Lat_Index：偏侧化指数，计算公式为LI=（左眼注视时长-右眼注视时长）/（总注视时长（左眼+右眼+双眼注视）） Binocular_Look_Perc：在测试场地内，双眼可见时长占总观测时长的百分比（对应GoPro拍摄画面中屏幕处于双眼视野范围的情况） Left_Gaze_Perc：在测试场地内，左眼可见时长占总观测时长的百分比（对应GoPro拍摄画面中屏幕处于左眼视野范围的情况） Right_Gaze_Perc：在测试场地内，右眼可见时长占总观测时长的百分比（对应GoPro拍摄画面中屏幕处于右眼视野范围的情况） No_Look_Perc：在测试场地内，双眼均不可见时长占总观测时长的百分比（对应GoPro拍摄画面中屏幕未处于其视野范围的情况） Look_Perc：在测试场地内，至少单眼可见时长占总观测时长的百分比（对应GoPro拍摄画面中屏幕处于其视野范围的情况） Z1_Pres：是否处于靠近屏幕的区域，编码为"Yes"（是）或"No"（否）