兽研1. 引言
在人类世，猫的角色已不再局限于家庭、花园和船只中的伴侣和捕鼠者[1]。家猫（Felis catus）日益被视为全球众多生态系统中生物多样性丧失的驱动因素（澳大利亚[2]、加拿大[3]和中国[4]）。数亿只自由放养的宠物猫和野猫通过捕食对野生动物造成影响。根据Lepczyk等人[5]对全球猫食性的综述，猫是极端泛化的捕食者，捕食受保护关注物种的比例高达16.7%（347/2084）。尽管猫在可获得的情况下主要捕食鸟类、哺乳动物和爬行动物（约90%[5]），但它们也可能形成个体捕食偏好并专食稀缺、脆弱的猎物[6]。
事实上，猫与63个物种的灭绝有关，并与啮齿动物一起位列全球威胁鸟类的顶级哺乳动物入侵捕食者之列[7]。但猫不仅捕食野生动物，它们还是病原体向野生动物溢出传播的被忽视的风险[8]，可通过杂交减少野猫的基因库（如欧洲野猫Felis silvestris[9]），并与野生食肉动物竞争食物[10]。猫还可能因野生动物仅仅恐惧其存在而间接降低野生动物的活动[11]。例如，鸟类的繁殖力可因亲鸟食物负载减少而降低[12]。
由于猫与人类长期建立的关系，自由放养宠物猫的管理与野生或野化食肉动物的管理存在根本差异。由于家猫是受人类照管的伴侣动物，限制其户外活动的策略往往受到主人观念的制约，主人认为猫应自由游荡并表达其天然捕猎行为[13,14]。在这些态度持续存在的同时，需要更好地理解影响猫户外行为的因素，以为有效管理策略提供依据。传统上，这些因素被归类为环境、内在和饲养相关决定因素。
猫户外行为的环境预测因素包括城市化程度。较低的居住密度增加了猫的家域范围[15,16]，但参见[17]。农村猫主人报告的捕获猎物频率也高于城市主人，而树木的存在显著增加了对鸟类的捕食率[18]。关于猫户外行为的内在预测因素，一项涉及25项研究的荟萃分析[15]以及对来自六个国家的875只猫的GPS追踪[19]显示，年轻的雄性猫具有更大的家域范围。然而，性别的影响并不总是明显[16,20]。饲养相关驱动因素表明未绝育猫[21]和未绝育雄性猫活动范围更远[19]具有显著影响，但该效应并不确定[15]。提供高蛋白食物减少了自由放养猫带回家的猎物数量[22]；同样，每日饲喂商品粮或自制食物而非麦麸或家庭残羹的猫，其粪便中的野生脊椎动物较少[23]。主人是否允许完全户外活动或在夜间限制猫的活动对英国的72只猫有强烈影响，不受限制的猫家域范围大75%[24]，但参见[16]。最后，家庭中是否有儿童或是否有狗存在也增加了猫的自由游荡[25]。
尽管上述部分驱动因素反映了文献中的相同趋势，但仍存在相当大的变异性，例如家域研究中的“离群猫”所反映的那样[17,19]。这种剩余的变异性是否可归因于猫的个性特征是一个新兴的研究领域。例如，在外向性（如活跃、警觉、好奇特质）上得分高而在神经质（如不安全感、焦虑、怕人）上得分低——猫科动物五种个性特征中的两种[26]——与更多猎物被带回住所相关[24,27]。另一方面，在宜人性（如亲昵、对人友好、温顺）和神经质上水平高的猫被发现家域范围较小且捕猎倾向较低[28]。尽管有这些令人鼓舞的新研究，一个被系统忽视的、解释个体变异性的驱动因素是猫与主人的关系。
有证据表明，猫与主人之间关系的质量可能影响主人对猫管理方式的决策。例如，将猫视为“孩子”或“最好的朋友”而非仅仅视为“宠物”的主人，更不可能让猫到户外活动[29]。同样，以频繁的猫主人互动为特征的“共生依赖”关系（例如，主人在喂食时通常陪伴在猫身边）的主人不会允许猫到户外活动，这与“随意”的猫主人关系形成对比，在后者中猫较少寻求与主人接近[30]。猫主人关系的另一个指标是猫与主人之间玩耍的频率[31,32]。值得注意的是，Cecchetti等人[22]发现，与猫进行仅5-10分钟的物体玩耍可使猫带回家的猎物减少25%。这一发现与Hall和Bradshaw[33]的实验研究一致，他们证明物体玩耍和捕食行为在动机上相关联，表明物体玩耍可能部分满足猫的捕猎动机。基于这些证据，我们旨在研究猫主人整体关系是否与猫的户外行为相关，正如最近对狗所描述的那样[34]。
为了记录猫的行为，我们决定使用动物携带式摄像机，这是一项卓越的技术，自Loyd等人[35]于2013年发表其里程碑式研究以来，该技术为猫的捕猎行为带来了新的见解。该研究表明，基于带回家猎物的捕食估计大大低估了猫的成功捕猎，正如视频记录所证明的那样[35]。此后，动物携带式摄像机不仅被用于记录捕食本身，还被用于研究捕食的决定因素。因此，摄像机数据显示，澳大利亚的野猫在开阔栖息地中的捕猎能力远优于在茂密栖息地中[36]。同样，法国宠物猫的捕猎频率取决于主人赋予的性格特征[28]。猫携带式摄像机甚至记录了在中非乍得可能与猫寄生虫传播相关的人类行为[37]，突显了该方法可应用的多样研究领域。据我们所知，我们的研究首次在智利使用动物携带式摄像机，该国拥有高度生物多样性的区域，自由放养猫正受到越来越多的科学关注，例如关于其运动生态学[38]和主人喂养昆虫饲料的意愿[39]。具体而言，我们在一个缺乏原生陆生捕食者的近乎原始岛屿上开展工作[40]，这对猫与野生动物的冲突而言是一个特别具有挑战性的环境[5,41]。
在本研究中，我们明确询问：(1) 智利最南部的户外猫会出现哪些行为；(2) 在人为-自然的环境梯度上，猫在何处花费更多时间；(3) 哪些因素（内在因素、饲养相关因素和猫主人关系）决定了户外猫的探索和捕猎相关行为。我们预测，与猫关系更强的主人，其猫较少参与猫与野生动物的冲突，表现出较少的探索和捕猎行为（一起玩耍降低捕食率[22]）。这些行为在老年猫和无依赖幼猫的雌猫中也会较少出现（雄性和年轻猫游荡更远[15,19]），在主人提供更高水平照料和监管的猫中也是如此（商业食品降低野生猎物的消费[23]，受限的猫游荡较少[24]）。我们努力使人们更加关注主人通过互动和情感纽带在塑造猫的户外行为中所起的作用。通过这种方式，可以将一个额外的维度纳入猫管理活动，即促进牢固的猫主人关系。此类活动在猫活动于野生栖息地附近的地方尤为重要。
2. 材料与方法
2.1. 研究区域
我们在纳瓦里诺岛（2,528 km²）开展工作，该岛位于智利最南部（南纬55°）。纳瓦里诺岛是合恩角生物圈保护区的一部分，以其亚南极景观的近原始特征而著称，这些景观由南方山毛榉林（Nothofagus spp.）、麦哲伦苔原（Sphagnum spp.）、高安第斯栖息地、湖泊和河流组成。岛上唯一的较大城镇是威廉斯港，约有2,200名居民（图1）。此外，岛屿北海岸沿线分布着几个小型农场，岛屿东部有一个小渔镇。引入的捕食者包括美洲水貂（Neogale vison）、家犬（Canis familiaris）和家猫。据主人报告，一些自由放养的猫会消失长达七天，猫也会走失（10年间12只猫）[42]。总体而言，在威廉斯港，作为宠物饲养的狗多于猫（215户中分别为40%和14%），大多数猫为本地来源（67%，n=30）[42]。
2.2. 参与研究的猫
共有12只家猫参与了我们的研究（2017/18年12月至4月及2018/19年同期）。通过邀请当地研究人员和兽医认识的、可能感兴趣的猫主人，以及走访花园中有猫的家庭来招募猫。仅选取经兽医在现场访问期间检查健康的猫、年龄≥1岁的成年猫以及每天至少在户外活动两小时的猫进入研究。在第一次佩戴摄像设备时表现出应激迹象的猫被排除在研究之外（n=3，来自15只入选猫）。十只参与研究的猫来自Puerto Williams的家庭，而两只猫住在镇外。
2.3. 猫-主人关系量表
我们使用Howell等人[31]开发的猫-主人关系量表（CORS）从主人角度评估猫与主人之间关系的强度（原始CORS经翻译（为西班牙语）并稍作修改的版本见补充材料表S1）。该量表是将广泛使用的Monash犬-主人关系量表[43]改编用于猫，分为三个分量表：（1）宠物-主人互动分量表（简称：互动）评估主人与猫玩耍、抚摸或交谈等活动的频率；（2）感知情感亲密度分量表（简称：情感）评估主人对“我的宠物帮助我度过艰难时期”等陈述的认同程度；（3）感知成本分量表（简称：成本）评估主人感受到养猫负担的强度。我们遵循Howell等人[31]的评分说明，为每个分量表计算1至5分的得分；得分越高，感知关系质量越好，即互动频繁、情感亲密度高、负担水平低。我们使用Cronbach's alpha确定各分量表的内部一致性，以>0.7作为可接受的一致性水平。
2.4. 猫-主人人口统计学数据
我们还使用了一份改编自研究区域先前研究[42]的问卷来收集猫和主人的背景数据。该问卷包含20个封闭式和开放式问题，涉及猫的背景、饲养管理、活动限制、观察到的猫-野生动物互动以及主人的个人信息（表A1）。
2.5. 动物携带式摄像机
我们为12只猫佩戴了专为家猫设计的小型摄像机（以下简称摄像器；Eyenimal；NUM'AXES公司，法国Olivet），安装在商用猫项圈上（图2a）。监测在两个夏季进行（总工作量见表1）。摄像器和项圈重约40 g，不足家猫平均体重的1%。它可录音并配备红外灯以便在黑暗时段拍摄。我们的目标是每只猫至少记录3小时。由于摄像器电池充满电后最多持续约3小时，我们估计至少需要对猫主人进行两次会话或访问。会话在一天中的不同时间进行，最早晨间会话从早上6点开始，最晚夜间会话从凌晨1点开始。然而，具体安排最终取决于主人的可用性和猫的户外活动时间表。请注意，猫在户外开始视频会话后可能再次进入家中（或附近另一房屋）。理想情况下，每只猫的监测在4天至2周内完成。每次会话平均持续2.05 ± 0.52小时（范围：0.09–5.33，n=52）。由于我们关注猫-野生动物互动，设备使用运动检测触发录制模式。摄像机工作时会发出蓝光，这可能对猫-动物互动产生影响。
2.6. 猫行为分析
在视频分析中，我们使用了Huck & Watson[44]的“猫摄像行为谱”，他们将Stanton等人[45]详细的“猫科动物标准化行为谱”改编用于猫摄像影像。在猫摄像行为谱的36种行为中，我们的视频影像中出现了25种行为。此外，我们记录了[45]中提到的呼噜声以及与主体或物体的互动（自行分类），共计28种行为（我们的行为谱详细描述见表A2）。根据每种行为的持续时间，我们将短暂的瞬时动作记录为发生频率（即点状态行为），将较长的动作或姿势记录为发生频率和行为持续时间（即状态事件行为）。补充材料中提供了三个示例视频片段（视频S1–S3）。视频编码使用开源软件BORIS 7.9.4版[46]。
为熟悉猫摄像行为谱并建立一致的编码程序，一名实习学生最初对五只猫（F02、F03、F05、M01和M05）进行编码，而E.S.对三只猫（F01、M03和M05）进行编码。随后对编码程序和不确定之处进行讨论直至达成共识。基于此过程，最终编码员（N.R.T.）接受了培训，并由N.R.T.和E.S.独立对一只猫进行编码以完成校准。然而，我们未评估正式的观察者间信度，这是本研究的一个局限性。
为便于统计分析，我们根据各行为共同的功能作用，将个体行为归纳为更广泛的行为类别。为此，我们采用了Stanton等人[45]的分类方法，其提出了改编自文献的15种常见行为类别。我们的行为归属于Stanton等人分类中的13个类别；例外为攻击性和刻板行为，这些行为在我们的参与猫中未观察到。例如，探索行为类别定义为与探索环境或调查特定刺激相关的调查行为。摄食类别指与获取和消费食物相关的任何行为。每种行为可归属于多个类别。总体而言，大多数行为归属于活跃类别，其次是平静、摄食、运动、不活跃和维持类别（表A2）。
此外，我们总结了猫在（i）房屋内（主人的家或其他住所）和城市环境中花费的时间，区分（ii）灰色城市环境，即缺乏主要植被，例如房屋下方、棚屋内、街道上或后院，以及（iii）绿色城市环境，即花园和公共绿地，以及自然栖息地如（iv）灌木林和（v）森林（图2b–f）。猫的家到Puerto Williams城市区域边界（可进入原生森林）的平均距离为80 ± 122.9米（中位数：35，范围：5–450）；因此，我们推测猫可能访问了自然环境，与其家庭位置无关。
2.7. 统计分析
我们使用箱线图总结猫在各类别中的整体行为。为检查狩猎相关行为（即觅食、狩猎、潜行和凝视—狩猎）和探索相关行为（探索、互动—物体、互动—社交和观看）在个体间是否存在显著差异，我们进行了Kruskal–Wallis检验。每只猫在上述五种环境（即室内、灰色城市、绿色城市、灌木林和森林环境）中花费的时间百分比绘制在条形图中。补充地，我们进行了主成分分析（PCA），以揭示猫是单独使用这五种环境还是最终可以分组（例如，按性别）。由于每只猫的环境空间使用由总计100%的比例表示（即一个栖息地的增加自动意味着另一个的减少），这些变量不独立。为满足PCA的独立性假设，我们因此应用了中心对数比变换，使数据适合PCA。
To understand which factors might influence the cat’s outdoor behavior, we used non-parametric univariate tests in all combinations (n = 40) for the five response variables (i.e., owner observation of hunting/harassment, hunting- and exploratory-related behaviors, and time spent in nature, as recorded by videocams) and eight explanatory variables (i.e., biological cat characteristics, husbandry-related characteristics, and the cat–owner relationship questionnaire; description in Table 2). Additionally, we checked whether the response variables were related to one another (10 combinations). To test whether there were relationships between two categorical variables, we used Fisher exact tests (n = 11); for the comparison of distributions between a continuous and a categorical variable, we used Wilcoxon rank sum tests for independent samples (n = 27); and to assess the strength of the relationship between two continuous variables, we used Spearman correlations (n = 12). We corrected multiple comparisons using the Holm–Bonferroni procedure within each family of response variables, correcting tests across the eight explanatory variables. We also corrected the ten tests among the response variables themselves. Given our low sample size (12 individuals), inferential tests may have limited the power to detect underlying trends in the data. Therefore, we computed effect sizes as rank–biserial correlations, obtained from the Wilcoxon Z statistic (
[内容请查看原文]
3. 结果
3.1. 猫的背景
12只参与猫中一半为雌性，一半为雄性；平均年龄为3.7 ± 3.1岁（范围：1–13）。大多数猫（75%，9/12）出生在Navarino岛，大多数（83.3%，10/12）作为伴侣饲养。参与猫的饲养标准相对较高。总计91.7%（11/12）的猫已绝育；83.3%（10/12）在过去一年内接种了狂犬病疫苗；75%（9/12）定期接受寄生虫治疗，但其中六只已到期，即自上次医疗处理已超过3个月。猫通常喂食干粮（91.7%，11/12），只有两只猫在多个家庭进食。
据主人反映，12只猫中仅有2只大部分时间受到限制，2只猫在夜间进入室内，其余猫（66.7%）白天和夜间均可自由活动。关于猫报告的户外行为，66.7%（8/12）的猫曾将动物带回家；其中大多数带回了鸟类（7/12）和鼠类（5/12）。提到的物种包括南鹦哥（Enicognathus ferrugineus）和棕领雀鹀（Zonotrichia capensis）。然而，主人观察到大多数猫有捕猎行为（83.3%，10/12）。同样，鸟类是最常被骚扰的动物之一（12只猫中有7只），其次是鼠类（4/12），以及水貂（1/12）和苍蝇（1/12）。两名参与者提到其猫曾带回家兔子并捕猎蜥蜴。最后，12只猫中有5只（41.7%）曾走失，走失时间在1至30天之间（7.8 ± 12.4）。
3.2. 猫与主人的背景信息
猫主人中67%为女性，33%为男性（n = 12）；其中一半拥有大学学位，25%完成了高中学业，16.7%完成了小学学业，0.8%完成了技术培训。参与者的平均年龄为42.5 ± 14.7岁（范围：26–78岁），他们在纳瓦里诺岛的平均居住时间为12.1 ± 15年（范围：2.5–45年）。总体而言，参与者拥有1.2 ± 0.4只猫（范围：1–2只）和0.7 ± 1只狗（范围：0–3只）。
3.3. 猫与主人的关系
CORS互动分量表的Cronbach's α为0.60，情感分量表为0.84，成本分量表为0.61。从互动分量表中删除第26题后，α升至可接受的0.74水平。由于成本分量表缺乏内部一致性，我们将其从分析中删除。所有参与者均表示与猫有高水平的互动（均值 = 4.7 ± 0.5，范围：3.4–5；量表评分1–5），问卷反映出中等程度的情感纽带（均值 = 3.8 ± 0.7，范围：2.4–4.6；量表评分1–5）。
3.4. 猫的户外行为
猫总共被监测了106小时24分钟，我们从中分析了总计50小时33分钟的摄像头录像，期间猫处于活动状态且摄像头正在录制。由于猫在观察到的行为总数和从事行为的总时间方面存在差异，所有数据均以百分比表示，频率使用每只猫观察到的行为总数计算，持续时间使用总观察时间计算。最常表现的行为（所有猫所有行为的频率，图3a）与探索行为相关，包括探索、互动—物体、互动—社交和观察（均值 = 8.5 ± 15.1，中位数 = 0.4，范围：0–45.3）。发声类别，代表呼噜声和发声行为，是第二常表现的行为（均值 = 6.4 ± 12.5，中位数 = 1.3，范围：0–45.8）；运动（接近、攀爬、跳跃、奔跑、潜行和行走行为）占第三位（均值 = 5.5 ± 8.8，中位数 = 1.1，范围：0–33.7）。进食类别，连同饮水、吃食、觅食、捕猎、潜行和凝视—捕猎行为，在猫中表现相对较少（均值 = 0.5 ± 0.7，中位数 = 0.2，范围：0–3.9）。关于在不同行为类别中花费的时间（仅所有猫的状态事件行为，图3b），猫同样在探索行为上花费最多时间（均值 = 15.5 ± 21.9，中位数 = 0.1，范围：0–66.1），其次是恐惧（仅梳理行为；均值 = 8.0 ± 8.2，中位数 = 5.6，范围：1.5–29.8）和非活跃行为（梳理、躺卧、警觉休息、睡眠、伸展行为；均值 = 5.7 ± 8.8，中位数 = 2.3，范围：1.5–39.4）。
关于捕猎相关行为（觅食、捕猎、潜行和凝视—捕猎），Kruskal–Wallis检验显著（H(11) = 30.96，p = 0.001），表明个体间存在高度差异（图4a）。一只乡村猫（M03）的捕猎值最高（均值 = 0.73 ± 0.65，中位数 = 0.55，范围：0.18–1.7）。未能拍摄到成功的捕猎事件；仅观察到M06在后院嗅闻一只已死亡的鸟，鉴定为棕领雀鹀（补充材料视频S1）。然而，探索相关行为（即探索、互动—物体、互动—社交和观察行为）在个体间无显著差异（H(11) = 2.41，p = 1.0；图4b）。
猫在花费时间的环境方面存在个体偏好（图5a）。总体而言，它们大部分时间花费在无植被的城市区域（灰色城市）（均值 = 2:14 ± 1:40 时:分，中位数 = 2:21，范围：0:08–5:44），即占其被监测总时间的2.4%至92.2%。猫也在室内花费大量时间（均值 = 1:3 ± 1:13 时:分，中位数 = 1:21，范围：0:04–3:11；即占被监测总时间的7.8%–80.2%）。猫是否频繁光顾自然栖息地（即灌木地和森林）因个体而异（均值 = 0:37 ± 0:55 时:分，中位数 = 0:08，范围：0:00–2:25；即占被监测总时间的0.0%–40.9%）。PCA的前两个成分共同解释了86.5%的方差，并在排序空间中揭示了两个主要的个体聚类（图5b）：四只个体（F02、F03、M01和M04）不频繁光顾自然区域，而另外四只猫（M02、M03、F05和F06）使用了所有五种类型的环境。其余个体（F01、F04、M05和M06）位于这些聚类之外，表明其中间行为特征。
3.5. 猫户外行为的预测因素
在对响应变量和解释变量进行的40项非参数检验（Fisher精确检验、Wilcoxon秩和检验和Spearman相关性检验）中（表2），经Holm–Bonferroni校正后，没有一项检验显示出显著关系；然而，我们检测到三个具有大效应量的视觉趋势[47]。据报道，老年猫比年轻猫捕猎更多（U = 4，p = 0.328；rrb = −0.75，95% CI：[−0.93，−0.25]；图6a）。CORS载荷较高的主人似乎比载荷较低的主人更多地观察到其猫骚扰鸟类或哺乳动物（U = 4.5，p = 0.904；rrb = −0.67，95% CI：[−0.92，−0.02]；图6b），并且主人报告参与捕猎的猫表现出更高水平的探索相关行为（U = 6，p = 0.981；rrb = −0.62，95% CI：[−0.90，−0.01]；图6c）。
关于五个响应变量本身之间的关系，10项非参数检验均不显著。然而，我们检测到一个趋势，即在自然栖息地停留时间多于城市栖息地或室内的猫，其主人观察到捕猎更多（U = 5，p = 0.64；rrb = −0.69，95% CI：[−0.92，−0.12]；图6d）。这些猫（在自然环境中停留更多时间）也表现出更高百分比的捕猎相关行为；这两个变量呈中度相关，rs (10) = 0.37（p = 1.0）。
4. 讨论
本研究记录了智利南部一个被近乎原始亚南极生态系统环绕的小镇中12只宠物猫的详细户外行为。我们的主要发现表明，猫与野生动物互动相关的行为更可能是个体变异的结果，而非由内在因素、饲养相关因素或猫–主人关系强度所驱动。重要的是，在50.5小时的监测期间未记录到主动捕食事件。因此，探索行为和捕猎行为只能在恐惧景观框架内进行解释[11,12]。
与我们的预测不同，我们没有发现猫–主人关系与猫的视频证据所显示的探索和捕猎行为之间存在负相关的证据；内在因素或饲养相关因素也不重要。然而，我们检测到两个趋势。一方面，猫–主人关系量表得分较高的主人观察到其猫更多的骚扰事件；另一方面，主人观察到捕猎更多的猫将更多时间分配给自然环境而非城市环境。这是一个有趣的发现，因为主人对猫的关注增加——更亲密关系的产物——提高了他们对猫–野生动物冲突的认识。事实上，认识到猫是不受欢迎的捕食者应激励一些主人承担管理其猫行为的责任[50]。例如，在一个为期4周的参与式公民科学项目中，猫主人在记录其猫的猎物后，对“家猫捕杀野生动物是一个严重问题”这一陈述的回答选项从“不同意”显著变为“同意”[51]。然而，更广泛的证据表明，捕猎被广泛视为猫的自然行为[50]，并且对猫影响野生动物的担忧并不影响主人限制其猫活动的决定[14,52]，而对猫安全的担忧则有影响[25,53,54]。值得注意的是，对猫安全的担忧反映在CORS中，其中包括“当你的猫死亡时，你认为对你来说会有多痛苦？”这一条目[31]。与此一致，对猫有更强情感依恋的主人已被证明比依恋较弱的主人更可能限制其猫的户外活动[29,30]。因此，如果猫限制运动强调与猫安全和主人–猫纽带相关的信息，特别是当对野生动物的担忧不足以激励行为改变时，可能会更有效[54]。
在这项探索性研究中，我们强调了猫–主人关系作为猫–野生动物冲突的驱动因素，但需要进一步研究。此类研究可以调查CORS[31]对自由放养猫家域的影响，或猫–主人纽带强度[30]对猫捕获猎物的影响。重要的是，猫对纽带的视角可能不一定反映基于问卷的主人的评估。因此，诸如陌生情境测试[55]之类的实验性评估将是理想的额外测试。该测试最初通过记录一岁幼儿在陌生人和母亲在场和缺席时的探索、寻求接近或接触维持行为来评估婴儿与母亲之间的依恋，并已被改编用于宠物–主人配对，以评估压力情境下的纽带关系（参见[34]用于狗和[56,57]用于猫）。当然，应测试相当数量的个体，因为猫–野生动物文献指向自由放养猫相当特殊的行为（例如，[6,28]）。
狩猎相关行为频率的个体差异确实是我们研究中一个高度显著的结果（图4a）。经Cecchetti等人[58]综述，猫狩猎行为的差异可归因于由供给食物类型驱动的个体营养需求[22]，但也与喂养方案有关（此处未评估）。例如，自由获取食物并以每日多次少量进食的方式摄入，可能模拟了野外的进食模式[59]，从而降低狩猎频率[22]。早期生活经历（如母猫的教导）也确实形成了个体化的狩猎行为[58]。最近，某些个性特征与狩猎活动相关联，活跃、好奇、好探究的猫在外向性特质集群中得分较高[27,28]。
猫对环境的偏好似乎也是由个体决定的。尽管我们通过PCA发现了两个聚类——偏好人工环境的猫与在从人工到更自然的环境梯度中活动的猫——但其他猫并不属于这两个聚类（图5）。据描述，猫的家域范围更接近1.5公顷（核密度估计均值[60]），而非10.9公顷的上限范围[61]，通常保持在距家60–100米的范围内[16,62]。鉴于这些活动范围，我们研究中选择家庭和城市区域的四只猫本应能够接触到森林栖息地，因为它们的家距离城市边界仅66.3 ± 48.5米（中位数=65，范围：10–125）。猫在栖息地利用方面的可塑性此前已有描述。例如，在以色列最近一项对GPS追踪猫的研究中，沿道路、建筑物内以及开阔的城市、自然和农业区域游荡是非选择性的[63]。在澳大利亚的宠物猫中也观察到了同样的情况，它们通常对城市栖息地或植被空间没有偏好[64]。这与将猫描述为更多与人类基础设施而非自然区域相关联的其他研究形成对比[19,38,65]。
尽管如此，当宠物猫居住在环境敏感区域附近时，需要高度的保护意识，因为它们可能游荡进入这些区域[38]，并以郊区猫两倍以上的速率带回猎物[66]。然而，对于原生植被碎片并未证实这一点[60]，这使我们认为猫游荡进入绿地的原因可能不仅仅是捕食。我们意识到GPS追踪是明确评估猫栖息地利用的更合适方法（参见Bischof等人[62]和Jensen等人[67]对超过90只个体的扎实研究）。然而，我们的研究凸显了动物携带相机在实现小规模评估方面的潜力，而这些评估并非总能通过GPS数据实现[63]，从而有助于我们理解猫在不同环境中的实际行为。
与狩猎行为和对城市或更自然环境环境的偏好不同，猫的探索模式相对一致，但值得关注，因为它们是表现最频繁的行为（考虑均值，分别以点事件和状态事件行为测量；图3）。与更大胆且更相关的捕食行为相比，这些行为在猫的行为研究中描述较少。然而，如果我们更好地理解探索与狩猎之间的关系，我们可能会更好地理解猫的捕食行为。例如，探索增加了获得猎物经验的机会，而这些经验在幼猫期获得后，可 later 影响狩猎偏好和技能[68]。事实上，我们检测到的趋势之一显示，主人报告有狩猎行为的猫表现出更高水平的探索行为。
最后，我们发现了一个趋势，即主人报告老年猫狩猎更多，这一结果与早期研究报告年轻猫猎物带回率更高的结果不一致[60,69]。重要的是，通过视频记录证实的真实捕食率显著高于主人报告的带回率[35]。反过来，猫是否带回猎物可能取决于特定的猫特征或其与主人的关系，因此这一捕食测量指标存在一定的不确定性。例如，在我们的研究中，主人对猎物的观察可靠性值得质疑，因为两名参与者报告的兔子和蜥蜴实际上在岛上并不存在[40]。不过，兔子可能作为逃逸宠物暂时存在于威廉斯港周边地区（Elke Schüttler，个人观察）。主人报告的狩猎活动与相机影像中与狩猎相关的行为之间也没有相关性；事实上，我们未能通过视频记录到任何活体捕食事件，这与其它研究的发现形成对比（例如，新西兰37只猫中23只有40次成功捕获[70]，美国55只猫中16只有39次[35]）。这使我们强调本研究的一些重要局限性。理想情况下，应监测超过12只个体并持续更长时间。然而，威廉斯港只有少数家庭（215户中的14%）养猫[42]；因此，招募受限。我们也无力购买技术先进的相机，如美国国家地理制作的KittyCams[35,36,70]；相反，我们使用了市售宠物相机，其仅能录制最长约3小时的片段，而KittyCams可录制10–12小时。这一不足可能降低了我们拍摄相对罕见的捕食事件的机会（流浪猫平均89.5%的时间不活动，仅0.9%的时间狩猎[71]）。然而，市售相机完成了其任务，因此对于开展科学研究面临资金困难的国家而言，是一种可负担的选择。
5. 结论
在这项探索性研究中，我们探讨了猫与野生动物冲突的一个相对较新的驱动因素，即猫与主人的关系。尽管我们未发现该驱动因素对猫与野生动物互动倾向的直接影响，但我们强调，报告与猫关系更密切的主人更频繁地观察到其猫干扰野生动物。因此，如果主人更加关注其猫——作为牢固关系的一部分——他们就会更加意识到其猫是野生动物的捕食者。鉴于猫主人对狩猎的看法多样，这并不必然意味着主人会为其猫承担责任[50]。然而，如果针对猫主人的行为改变倡议能够处理看似冲突的价值观，如猫的福利与户外限制，猫的圈养可能会逐步实现[50,54]。由于猫在很大程度上表现为个体行为，我们建议猫主人仔细观察其猫，甚至可像本研究中一样使用商用摄像机，以识别可能的猎手并承担责任。这一责任应从夜间禁闭[24,54,72]、使用项圈铃铛或彩色袖套[73,74]以及增加猫与主人的互动[30]（如玩耍[22]）开始，作为促进猫与野生动物共存的初步措施。
查看英文原文▾
1. Introduction
In the Anthropocene, the role of cats is no longer confined to that of companions and mousers in homes, gardens, and ships [1]. Domestic cats (Felis catus) are increasingly recognized as agents of biodiversity loss in numerous ecosystems around the globe (in Australia [2], in Canada [3], and in China [4]). Hundreds of millions of free-ranging pet and feral cats impact wildlife through predation. According to the synthesis of Lepczyk et al. [5] on global cat diet, cats are extreme generalist predators, preying up to 16.7% (347/2084) on species of conservation concern. Although cats predominantly consume birds, mammals, and reptiles upon availability (~90% [5]), they can also develop individual hunting preferences and specialize on scarce, vulnerable prey [6].
Indeed, cats are related to 63 species extinctions and—together with rodents—figure among the top mammalian invasive predators worldwide endangering birds [7]. But not only do cats hunt wildlife, they are also an overlooked risk for pathogen spillover to wildlife [8], can reduce the gene pool of wildcats through hybridization (e.g., the European wildcat Felis silvestris [9]), and compete with wild carnivores for food [10]. Cats may also indirectly reduce wildlife activity as a result of fear of their mere presence [11]. For example, bird fecundity can be diminished through a reduction in parental food load [12].
Due to the long-established relation of cats with humans, the management of free-ranging pet cats differs fundamentally from that of wild or feral carnivores. Because domestic cats are companion animals under human care, strategies to limit their outdoor access are often constrained by owners’ beliefs that cats should be free to roam and express their natural hunting behavior [13,14]. While these attitudes persist, a better understanding of the factors influencing cats’ behavior outdoors is needed to inform effective management strategies. Traditionally, these factors have been classified into environmental, intrinsic, and husbandry-related determinants.
Among the environmental predictors of cat outdoor behavior is the degree of urbanization. A lower residential density increased the home ranges of cats [15,16], but see [17]. Rural cat owners also reported higher frequencies of captured prey than urban owners, while the presence of trees significantly increased rates of predation on birds [18]. Regarding the intrinsic predictors of cat outdoor behavior, younger male cats have been found to have larger home ranges, as revealed by a meta-analysis involving 25 studies [15] and by GPS tracking of 875 cats from six countries [19]. The influence of sex was not always evident, though [16,20]. Husbandry-related drivers indicate a significant effect of intact cats [21] and intact males moving further [19], but this effect was not conclusive [15]. Providing high-protein food reduced the number of prey brought home by free-ranging cats [22]; similarly, cats fed with commercial or prepared food on a daily basis versus wheat bran or household scraps had fewer wild vertebrates in their scats [23]. Whether owners allowed full outdoor access or restricted cat movement during the night had a strong effect on 72 cats in Great Britain, with 75% larger home ranges in unrestricted cats [24], but see [16]. Finally, whether the household included children or whether dogs were present also increased free roaming in cats [25].
Although some of the forementioned drivers reflect the same tendency in the literature, there is still a considerable amount of variability at stake, reflected, for example, by “outlier cats” in home range studies [17,19]. Whether this remaining variability can be ascribed to personality traits in cats is an emerging field of research. For instance, high scores for extraversion (e.g., active, vigilant, curious traits) and low ones for neuroticism (e.g., insecure, anxious, fearful of people), two of the five personality traits in felids [26], were associated with more prey returned to home [24,27]. On the other hand, high levels of agreeableness (e.g., affectionate, friendly to people, gentle) and neuroticism were found in cats with smaller range sizes and less hunting propensity [28]. Despite these encouraging new studies, one systematically overlooked driver accounting for individual variability is the cat–owner relationship.
There is evidence that the quality of the cat–owner relationship may influence owners’ decisions regarding cat management. For example, owners who perceived their cat as a “child” or “best friend,” rather than simply as a “pet,” were less likely to provide outdoor access [29]. Similarly, owners characterized by a “co-dependent” relationship, shaped by frequent cat–owner interactions (e.g., the owner typically remains with the cat during feeding), would not allow their cats outdoor access, in contrast to “casual” cat–owner relationships, in which cats seek less proximity to their owners [30]. Another indicator of the cat–owner relationship is the frequency of play between cats and their owners [31,32]. Notably, Cecchetti et al. [22] found that only 5–10 min of object play with cats reduced prey brought home by 25%. This finding is consistent with the experimental study by Hall and Bradshaw [33], who demonstrated that object play and predatory behavior are motivationally linked, suggesting that object play may partially satisfy cats’ motivation to hunt. Building on this evidence, we aimed to investigate whether the overall cat–owner relationship is associated with cats’ outdoor behavior, as has recently been described for dogs [34].
To document cat behavior, we decided to use animal-borne cameras, an outstanding technique, which has brought new insights into the hunting behavior of cats since Loyd et al. [35] published their landmark study in 2013. This study demonstrated that estimates of prey capture based on prey brought home substantially underestimate cats’ successful hunting, as evidenced by video recordings [35]. Thereafter, animal-borne video cameras have not only been used to document predation itself but also to investigate the determinants of predation. Thus, camera data showed that feral cats in Australia were much better hunters in open versus dense habitat [36]. Similarly, the hunting frequency of pet cats in France depended on owner-ascribed personality traits [28]. Cat-mounted cameras even documented human behaviors potentially associated with parasite transmission in cats in Chad, Central Africa [37], highlighting the diverse research areas in which this method can be applied. To our knowledge, our study uses animal-borne cameras for the first time in Chile, a country with highly biodiverse regions where free-ranging cats are receiving increasing scientific attention, e.g., regarding their movement ecology [38] and owner willingness to feed insect meals [39]. Specifically, we worked on a near-pristine island that lacks native terrestrial predators [40], a particularly challenging setting for cat–wildlife conflict [5,41].
In this study, we explicitly asked (1) which behaviors occur in outdoor cats in southernmost Chile, (2) where on an anthropogenic–natural environmental gradient do cats spend more time, and (3) which factors (intrinsic, husbandry-related, and cat–owner relationship) determine behaviors related to exploration and hunting behaviors in outdoor cats. We predicted that cats of owners indicating a stronger relationship with their cats would get less engaged in cat–wildlife conflict, exhibiting less exploratory and hunting behavior (playing together decreases predation rates [22]). Those behaviors would also be less present in older cats and female cats without dependent kittens (male and younger cats roam further [15,19]) and in cats whose owners provide higher levels of care and surveillance (commercial food decreases the consumption of wild prey [23], and restricted cats roam less [24]). We endeavor to draw more attention to the role owners play in shaping their cats’ outdoor behavior through their interactions and emotional bonds. In this way, an additional dimension could be incorporated into cat management campaigns, namely, the promotion of a strong cat–owner relationship. Such campaigns are particularly relevant wherever cats range near wild habitats.
2. Materials and Methods
2.1. Study Area
We worked on Navarino Island (2.528 km2), southernmost Chile (55° S). Navarino Island is part of the Cape Horn Biosphere Reserve, outstanding for its near-pristine character of sub-Antarctic landscapes composed of southern beech forests (Nothofagus spp.), Magellanic tundra (Sphagnum spp.), high-Andean habitats, lakes, and rivers. The only larger town on the island is Puerto Williams, with approximately 2.200 inhabitants (Figure 1). Additionally, several small-scale farms are located along the northern coast of the island, and there is a small fishing town in the east of the island. Introduced predators are American mink (Neogale vison), domestic dogs (Canis familiaris), and domestic cats. Some free-ranging cats have been reported by their owners to disappear for up to seven days, and cats also get lost (12 cats over a 10-year period) [42]. In general, there are more dogs than cats kept as pets in Puerto Williams (40% versus 14% of 215 households), and most cats are of local origin (67%, n = 30) [42].
2.2. Participating Cats
A total of 12 domestic cats participated in our study (December–April 2017/18 and 2018/19). Cats were recruited by inviting cat owners known by local researchers and veterinarians to be possibly interested and by approaching households with cats in their gardens. Only healthy cats, checked by a veterinarian during an in situ visit, adult cats aged ≥1 years, and cats that spent at least two hours outside each day were selected for the study. Cats showing signs of stress when wearing the camera device during their first session were excluded from the study (n = 3, from 15 selected cats). Ten participating cats were from households in Puerto Williams, while two cats lived outside the town.
2.3. Cat–Owner Relationship Scale
We used the Cat–Owner Relationship Scale (CORS) developed by Howell et al. [31] to assess the strength of the relationship between cats and their owners, from the owner’s perspective (see Table S1 of the Supplementary Materials for a translated (into Spanish) slightly modified version of the original CORS). The scale is an adaptation of the widely used Monash Dog–Owner Relationship Scale [43] to cats and is divided into three subscales: (1) The Pet–Owner Interaction subscale (in short: Interaction) evaluates how often the owner engages in activities such as playing, petting, or talking with the cat; (2) the Perceived Emotional Closeness subscale (in short: Emotion) evaluates the extent to which the owner agrees with statements such as “My pet helps me get through tough times”; and (3) the Perceived Costs subscale (in short: Costs) evaluates how strongly the owner feels the burden of having a cat. We followed the scoring instructions in Howell et al. [31] and calculated a score from 1 to 5 for each subscale; the higher the score, the better the quality of the perceived relationship, meaning frequent interaction, high levels of emotional closeness, and low levels of burden. We used Cronbach’s alpha to determine the internal consistency of each subscale, using a value of >0.7 as an acceptable level of consistency.
2.4. Demographic Cat–Owner Data
We also applied a questionnaire to collect background data on cats and owners, adapted from previous research in the study area [42]. This questionnaire contained 20 closed- and opened-ended questions on the cats’ background, husbandry, restriction of movement, and observed cat–wildlife interactions, as well as personal information of owners (Table A1).
2.5. Animal-Borne Cameras
We fitted 12 cats with small video cameras (in what follows, videocams) designed for domestic cats (Eyenimal; NUM’AXES company, Olivet, France) attached to commercial cat collars (Figure 2a). Monitoring took place during two summers (total effort in Table 1). The videocam and collar weighed ~40 g, which is <1% of the average body weight of a domestic cat. It recorded sound and was equipped with infrared light to film during dark hours. We aimed at recording a minimum of 3 h per cat. As the batteries of the videocam lasted a maximum of ~3 h once fully charged, we estimated a minimum effort of two sessions or visits to the cat owner. The sessions took place at different times of the day, 6 AM being the start of the earliest morning session and 1 AM that of the latest night session. However, scheduling ultimately depended on the availability of the owner and the outdoor schedule of cats. Note that cats could enter the home (or another house in the vicinity) again after starting the video session outside. Ideally, the monitoring of each cat was completed within a period of 4 days to 2 weeks. Each session lasted on average 2.05 ± 0.52 h (range: 0.09–5.33, n = 52). As we were interested in cat–wildlife interactions, the device was used in the mode to start recording when movement was detected. A blue light shone when cameras were working, which might have had an influence on cat–animal interactions.
2.6. Analysis of Cat Behaviors
For our video analysis we used the “catcam ethogram” of Huck & Watson [44], who adapted the detailed “standardized ethogram for the Felidae” by Stanton et al. [45] to cat-camera footage. Of the 36 behaviors in the catcam ethogram, 25 behaviors occurred in our video footage. Additionally, we recorded the behaviors purr, mentioned in [45], and interaction with subjects or objects (own classification), totaling 28 behaviors (detailed description of our ethogram in Table A2). Depending on the duration of each behavior, we recorded short instantaneous actions as frequencies of occurrence (i.e., point state behaviors) and longer actions or postures as both frequencies of occurrence and time the behavior persisted (i.e., state event behaviors). Three exemplary video excerpts are provided in the Supplementary Materials (Videos S1–S3). For video coding, we used the open-source software BORIS, version 7.9.4 [46].
To familiarize ourselves with the catcam ethogram and establish consistent coding procedures, an internship student initially coded five cats (F02, F03, F05, M01, and M05), while E.S. coded three cats (F01, M03, and M05). Coding procedures and uncertainties were subsequently discussed until consensus was reached. Based on this process, the final codifier (N.R.T.) was trained, and one cat was coded independently by N.R.T. and E.S. to finish the calibration. However, we did not assess formal interobserver reliability, which is a limitation of the study.
To facilitate statistical analyses, we condensed individual behaviors into broader behavioral categories according to their shared functional roles. For this purpose, we followed the categorization of Stanton et al. [45], who suggested 15 common behavioral categories adapted from the literature. Our behaviors fell into 13 of Stanton et al.’s categories; the exceptions were aggressive and stereotypic behaviors, which were not observed among our participating cats. The category exploratory behaviors, for instance, is defined by investigative behaviors related to the exploration of the environment or the investigation of specific stimuli. The category feeding refers to any behavior related to the acquisition and consumption of food. Each behavior could occur in several categories. Overall, most behaviors fell into the category active, followed by the categories calm, feeding, locomotion, inactive, and maintenance (Table A2).
Additionally, we summarized the time cats spent (i) inside a house (the owner’s home or another home) and in the urban environment, differentiating between (ii) grey urban, meaning lacking predominant vegetation, e.g., under a house, in a shed, on the street, or in a backyard, and (iii) green urban environments, meaning gardens and public green spaces, as well as natural habitats such as (iv) shrubland and (v) forest (Figure 2b–f). The mean distance of the cat’s home to the limit of the urban area of Puerto Williams, where native forest could be accessed, was 80 ± 122.9 m (median: 35, range: 5–450); therefore, we supposed that cats could have visited a natural environment independent of their home locations.
2.7. Statistical Analysis
We used box plots to summarize the cats’ overall behaviors across categories. To check whether hunting-related (i.e., foraging, hunting, stalking, and stare—hunting) and exploratory-related behaviors (exploring, interaction—objects, interaction—social, and watching) showed significant differences between individuals, we performed Kruskal–Wallis tests. The percentage of time spent in each of the five above-mentioned environments (i.e., inside, grey urban, green urban, shrubland, and forest environments) was drawn in bar plots for each cat. Complementarily, we performed a Principal Component Analysis (PCA) to reveal whether cats used these five environments individually or eventually could be grouped (e.g., by their sex). Since each cat’s environmental space use was represented by proportions that together totaled 100% (i.e., an increase in one habitat automatically means a decrease in another), these variables were not independent. To be able to meet the PCA assumption of independency, we therefore applied a centered log-ratio transformation to make the data suitable for PCA.
To understand which factors might influence the cat’s outdoor behavior, we used non-parametric univariate tests in all combinations (n = 40) for the five response variables (i.e., owner observation of hunting/harassment, hunting- and exploratory-related behaviors, and time spent in nature, as recorded by videocams) and eight explanatory variables (i.e., biological cat characteristics, husbandry-related characteristics, and the cat–owner relationship questionnaire; description in Table 2). Additionally, we checked whether the response variables were related to one another (10 combinations). To test whether there were relationships between two categorical variables, we used Fisher exact tests (n = 11); for the comparison of distributions between a continuous and a categorical variable, we used Wilcoxon rank sum tests for independent samples (n = 27); and to assess the strength of the relationship between two continuous variables, we used Spearman correlations (n = 12). We corrected multiple comparisons using the Holm–Bonferroni procedure within each family of response variables, correcting tests across the eight explanatory variables. We also corrected the ten tests among the response variables themselves. Given our low sample size (12 individuals), inferential tests may have limited the power to detect underlying trends in the data. Therefore, we computed effect sizes as rank–biserial correlations, obtained from the Wilcoxon Z statistic (
[内容请查看原文]
), and report their 95% confidence intervals to provide an estimate of the magnitude and uncertainty of the observed effects for visually detected trends. A large effect between groups is given with |rc| ≥ 0.43 (range: −1 to 1 [47]). All statistical analyses were carried out in the program R, version 4.1.2 [48]. We also declare the use of ChatGPT (version GPT-5.5, OpenAI, San Francisco, CA, USA) for orientation in R programming related to the PCA, the Holm–Bonferroni procedure, and the choice of effect sizes. All generated code was verified independently. Rank–biserial correlations were calculated using the R package “effectsize” [49]. Data are available as an Excel file in Table S2 of the Supplementary Materials.3. Results
3.1. Cat Background
Half of the 12 participating cats were female, half were male; the mean age was 3.7 ± 3.1 years (range: 1–13). The majority of the cats (75%, 9/12) were born on Navarino Island, and most (83.3%, 10/12) were held for companionship. The husbandry standard of the participating cats was relatively high. A total of 91.7% (11/12) of the cats were neutered; 83.3% (10/12) were vaccinated against rabies within the last year; and 75% (9/12) were regularly treated against parasites, but six of them were due, i.e., more than 3 months had passed since the last medical treatment. Cats were typically fed with dry food (91.7%, 11/12), and only two cats ate at more than one household.
According to their owners, only two out of the 12 cats were restricted most of the time, two cats entered their homes at night, and the remaining cats (66.7%) were able to roam freely day and night. Regarding the cat’s reported outdoor behavior, 66.7% (8/12) of the cats had brought home animals; the majority brought birds (7/12) and mice (5/12). Species mentioned were the Austral Parakeet (Enicognathus ferrugineus) and the Rufous-collared Sparrow (Zonotrichia capensis). However, owners observed most of their cats hunting (83.3%, 10/12). Again, birds were among the animals most frequently harassed (seven out of 12 cats), followed by mice (4/12), as well as mink (1/12) and flies (1/12). Two participants mentioned that their cats had brought home rabbits and hunted lizards. Finally, five out of the 12 cats (41.7%) had gone lost for spans between 1 and 30 days (7.8 ± 12.4).
3.2. Cat–Owner Background
Cat owners were 67% women and 33% men (n = 12); half of them had a university degree, 25% had finished high school, 16.7% had finished elementary school, and 0.8% had completed technical training. The participants’ mean age was 42.5 ± 14.7 years (range: 26–78), and their mean residence time on Navarino Island was 12.1 ± 15 years (range: 2.5–45). In total, the participants owed 1.2 ± 0.4 cats (range: 1–2) and 0.7 ± 1 dogs (range: 0–3).
3.3. Cat–Owner Relationship
Cronbach’s α was 0.60 for the Interaction subscale of the CORS, 0.84 for the Emotion subscale, and 0.61 for the Cost subscale. Dropping question 26 from the Interaction subscale raised α to an acceptable level of 0.74. As the Cost subscale lacked internal consistency, we deleted it from the analysis. All participants expressed high levels of interaction with their cats (mean = 4.7 ± 0.5, range: 3.4–5; scale scores were 1–5) and a moderate emotional bond was reflected by the questionnaires (mean = 3.8 ± 0.7, range: 2.4–4.6; scale scores were 1–5).
3.4. Cat Outdoor Behavior
Cats were monitored for a total of 106 h and 24 min, from which we analyzed a total of 50 h and 33 min of camera footage during which cats were active and cameras were recording. Because cats differed in both the total number of observed behaviors and the total time spent performing behaviors, all data were expressed as percentages, using each cat’s total observed behaviors for frequencies and total observation time for durations. The most often displayed behaviors (frequencies of all behaviors across all cats, Figure 3a) were related to exploratory behaviors, including exploring, interaction—objects, interaction—social, and watching (mean = 8.5 ± 15.1, median = 0.4, range: 0–45.3). The category vocalizations, representing the behaviors purring and vocalizing, was the second most displayed behavior (mean = 6.4 ± 12.5, median = 1.3, range: 0–45.8); locomotion (approaching, climbing, jumping, running, stalking, and walking behaviors) occupied the third place (mean = 5.5 ± 8.8, median = 1.1, range: 0–33.7). The category feeding, together with the behaviors drinking, eating, foraging, hunting, stalking, and stare—hunting, was relatively little displayed among cats (mean = 0.5 ± 0.7, median = 0.2, range: 0–3.9). Regarding time spent in the different behavioral categories (only state event behaviors across all cats, Figure 3b), cats also occupied most time in exploratory behaviors (mean = 15.5 ± 21.9, median = 0.1, range: 0–66.1), followed by fear (only behavior grooming; mean = 8.0 ± 8.2, median = 5.6, range: 1.5–29.8) and inactive behaviors (grooming, lying, resting alert, sleeping, stretching behaviors; mean = 5.7 ± 8.8, median = 2.3, range: 1.5–39.4).
Regarding the hunting-related behaviors (foraging, hunting, stalking, and stare—hunting) the Kruskal–Wallis test was significant (H(11) = 30.96, p = 0.001), indicating high individual differences between individuals (Figure 4a). A rural cat (M03) showed the greatest values for hunting (mean = 0.73 ± 0.65, median = 0.55, range: 0.18–1.7). No successful hunting event could be filmed; only sniffing by M06 of an already dead bird in the backyard, identified as a Rufous-collared Sparrow, was observed (Video S1 of the Supplementary Materials). However, the exploratory-related behaviors (i.e., exploring, interaction—objects, interaction—social, and watching behaviors) yielded no significant differences between individuals (H(11) = 2.41, p = 1.0; Figure 4b).
Cats had individual preferences in regard to which environment they spent their time in (Figure 5a). Overall, they spent most time in urban areas without vegetation (grey urban) (mean = 2:14 ± 1:40 hh:min, median = 2:21, range: 0:08–5:44), that is, between 2.4 and 92.2% of the total time they were monitored. Cats spent also much time inside a house (mean = 1:3 ± 1:13 hh:min, median = 1:21, range: 0:04–3:11; i.e., 7.8–80.2% of total time monitored). Whether cats frequented natural habitats (i.e., shrubland and forest) was individually different (mean = 0:37 ± 0:55 hh:min, median = 0:08, range: 0:00–2:25; i.e., 0.0–40.9% of total time monitored). The first two components of the PCA together explained 86.5% of the variance and revealed two main clusters of individuals in the ordination space (Figure 5b): Four individuals (F02, F03, M01, and M04) did not frequent natural areas, while another four cats (M02, M03, F05, and F06) used all five types of environments. The rest of the individuals (F01, F04, M05, and M06) were positioned outside these clusters, indicating intermediate behavioral profiles.
3.5. Predictors of Cat Outdoor Behavior
Of the 40 non-parametric tests (Fisher exact tests, Wilcoxon rank sum tests, and Spearman correlations) of response and explanatory variables (Table 2), no test yielded a significant relationship after the Holm–Bonferroni correction; however, we detected three visual trends with large effect sizes [47]. Older cats were reported to hunt more than younger cats (U = 4, p = 0.328; rrb = −0.75, 95% CI: [−0.93, −0.25]; Figure 6a). Owners with higher CORS loadings seemingly observed their cats’ harassing birds or mammals more than owners with lower loadings (U = 4.5, p = 0.904; rrb = −0.67, 95% CI: [−0.92, −0.02]; Figure 6b), and cats reported by their owners to engage in hunting showed higher levels of exploratory-related behaviors (U = 6, p = 0.981; rrb = −0.62, 95% CI: [−0.90, −0.01]; Figure 6c).
Regarding the relationships between the five response variables themselves, none of the 10 non-parametric tests was significant. Nevertheless, we detected a trend that cats that spent more time in natural habitats than in urban habitats or inside homes were observed by their owners to hunt more (U = 5, p = 0.64; rrb = −0.69, 95% CI: [−0.92, −0.12]; Figure 6d). Those cats (spending more time in nature) also showed higher percentages of hunting-related behaviors; these two variables were moderately correlated, rs (10) = 0.37 (p = 1.0).
4. Discussion
This study documented the detailed outdoor behavior of 12 pet cats in a small town surrounded by near-pristine sub-Antarctic ecosystems in southern Chile. Our main finding indicates that cats’ behaviors related to cat–wildlife interactions are more likely the result of individual variability than driven by intrinsic, husbandry-related factors or the strength of the cat–owner relationship. Importantly, no active predation event was recorded during 50.5 h of monitoring. Therefore, the exploratory and hunting behaviors can only be interpreted within the framework of the landscape of fear [11,12].
Different to our predictions, we found no evidence of a negative correlation between the cat–owner relationship and the cats’ video-evidenced exploration and hunting behaviors; neither did intrinsic or husbandry-related factors matter. However, we detected two trends. On the one hand, owners with higher scores on the Cat–Owner Relationship Scale observed more harassment incidents in their cats; on the other hand, cats whose owners observed them hunting more allocated more time to natural than to urban environments. This is an interesting finding, as increased attention of owners toward their cats—the product of a closer relationship—raised their awareness of cat–wildlife conflict. Indeed, the recognition of the cat as an undesired predator should incentivize some owners to assume responsibility for managing their cats’ behavior [50]. For example, cat owners significantly changed their answer options for the statement “Domestic cats killing wildlife is a serious problem” from “disagree” to “agree” after documenting their cat’s prey in the framework of a 4-week participatory Citizen Science project [51]. However, broader evidence suggests that hunting is widely regarded as a natural behavior in cats [50] and that concern about cats’ impacts on wildlife do not influence owners’ decisions to restrict their cats’ movements [14,52], whereas concerns about cats’ safety do [25,53,54]. Notably, concern for a cat’s safety is reflected in the CORS, which includes the item “How traumatic do you think it will be for you when your cat dies?” [31]. Consistent with this, owners with stronger emotional attachments to their cats have been shown to be more likely to restrict their cats’ outdoor access than less attached owners [29,30]. Therefore, cat containment campaigns may be more effective if they emphasize messages related to cats’ safety and the owner–cat bond, particularly when concern for wildlife is insufficient to motivate behavior change [54].
In this explorative study, we highlighted the cat–owner relationship as a driver of cat–wildlife conflict, but further research is needed. Such research could investigate the influence of the CORS [31] on home ranges of free-ranging cats or the strength of the cat–owner bond [30] on prey caught by cats. Importantly, the cat’s perspective on the bond might not necessarily reflect a questionnaire-based evaluation of the owner. Therefore, experimental assessments like the Strange Situation Test [55] would be an ideal additional test. This test originally evaluated attachment between infants and their mothers through the recording of exploratory, proximity-seeking, or contact-maintenance behaviors in one-year-olds in the presence and absence of a stranger and the mother and has been adapted to pet–owner dyads to evaluate bonding in a stressful situation (see [34] for dogs and [56,57] for cats). Definitely, a considerable number of individuals should be tested, as the cat–wildlife literature points to fairly idiosyncratic behavior in free-ranging cats (e.g., [6,28]).
Individual variability in the frequency of hunting-related behaviors was indeed a highly significant result in our study (Figure 4a). Reviewed by Cecchetti et al. [58], differences in the hunting behavior of cats can be ascribed to individual nutrient requirements driven by the type of provisioned food [22], but also the feeding scheme (here, not assessed). For example, free access to food ingested as several small daily meals might simulate feeding patterns in the wild [59], thereby reducing the frequency of hunting [22]. The early life experience (e.g., teaching by the mother) does also form individualistic hunting behaviors [58]. Recently, certain personality traits have been linked to hunting activity, with active, curious, inquisitive cats scoring high for the trait cluster extraversion [27,28].
A cat’s preference for its environment appears to be individually determined as well. Although we found two PCA clusters of cats preferring human-made surroundings versus cats spending their time along the gradient of human-made to more natural environments, other cats did not fit into these two clusters (Figure 5). Cats are described to have home ranges closer to 1.5 ha (mean kernel density estimate [60]) than to the upper range of 10.9 ha [61], staying normally as close as 60–100 m to their homes [16,62]. Given these movement ranges, the four cats that selected homes and urban areas in our study should have had access to forested habitat, as their homes were located only 66.3 ± 48.5 m from the urban boundary (median = 65, range: 10–125). Plasticity in habitat use has been previously described for cats. For example, in a recent study of GPS-tracked cats in Israel, roaming along roads, in buildings, and in open urban, natural, and agricultural areas was non-selective [63]. The same was observed in pet cats in Australia, which generally showed no preference for urban habitats or vegetated space [64]. This is in contrast to other studies describing cats as being more associated with human infrastructure than with natural areas [19,38,65].
Notwithstanding, a high level of conservation awareness is needed when pet cats reside near environmentally sensitive areas, as they might roam into these [38] and return prey at more than twice the rate of suburban cats [66]. This was not evidenced for fragments of native vegetation, though [60], which makes us think that cats might wander into green areas for reasons other than predation only. We are aware that GPS tracking is the more adequate method to explicitly assess habitat use in cats (see the solid studies with over 90 individuals by Bischof et al. [62] and Jensen et al. [67]). However, our study highlights the potential of animal-borne cameras to enable small-scale assessments that are not always feasible with GPS data [63], thereby contributing to our understanding of what cats actually do in different environments.
Different from hunting behaviors and preference for urban or more natural environmental settings, the cats’ exploration patterns were relatively uniform but deserve attention, as they were the most frequently displayed behaviors (considering the mean values, measured both as point events and for state event behaviors; Figure 3). Those behaviors have been less commonly described in behavioral studies of cats compared to the bolder and more relevant predation behaviors. However, if we better understand how exploration and hunting are related, we might better understand predation in cats. For example, exploration increases opportunities to acquire experience with prey, which, when gained during kittenhood, can later influence hunting preferences and skills [68]. Indeed, one of our detected trends showed that cats whose owners reported hunting exhibited higher levels of exploratory behaviors.
Finally, we found a trend that older cats were reported by their owners to hunt more, a result at odds with earlier studies reporting higher prey returns for younger cats [60,69]. Importantly, true predation rates as evidenced by video recording are significantly higher than owner-reported return rates [35]. In turn, whether a cat returns prey might depend on specific cat characteristics or its bond to the owner, thus leaving some uncertainty regarding this measurement of predation. In our study, for example, the reliability of owner observations on prey can be questioned, as rabbits and lizards reported by two participants are actually absent on the island [40]. Rabbits, though, might be temporarily present as escaped pets in the surroundings of Puerto Williams (Elke Schüttler, personal observation). There was also no correlation between owner-reported hunting activity and hunting-related behaviors in the camera footage; in fact, we could not video-record any live predation event, in contrast to findings from other studies (e.g., 40 successful captures in 23 out of 37 cats in New Zealand [70] and 39 in 16 out of 55 cats in the USA [35]). This makes us highlight some important limitations of our research. It would have been desirable to monitor more than 12 individuals and for longer time periods. However, few households (14% of 215) in Puerto Williams hold cats [42]; therefore, recruitment was limited. We also could not afford to buy technologically advanced cameras, such as the KittyCams produced by National Geographic, USA [35,36,70]; instead, we used commercially available pet cameras, which allowed recording of sessions up to ~3 h only, compared to the 10–12 h recording of KittyCams. This shortcoming might have reduced our chance to film the relatively rare predation events (stray cats spent on average 89.5% of their time inactive and only 0.9% hunting [71]). However, the commercially available cameras did their job and hence represent an affordable option for countries where conducting science is financially challenging.
5. Conclusions
In this explorative study, we have addressed a relatively new driver of cat–wildlife conflict, namely, the cat–owner relationship. Although we did not find a direct influence of this driver in the cat’s propensity to interact with wildlife, we highlight that owners reporting a closer relationship observed their cats disturbing wildlife more often. Hence, if owners pay more attention to their cats—as part of a strong relationship—they become more aware of their cats as predators of wildlife. Given the diversity of perceptions of hunting among cat owners, this does not necessarily mean that owners assume responsibility for their cats [50]. However, if behavioral change initiatives directed at cat owners address seemingly conflicting values, such as cat welfare and outdoor restriction, cat containment may be achieved gradually [50,54]. As cats largely behave individually, we recommend that cat owners carefully observe their cats, maybe even using commercial videocams, as in this study, to identify possible hunters and assume responsibility. This responsibility should start with night-time confinement [24,54,72], the use of collar-mounted bells or multicolored sleeves [73,74], and increased cat–owner interaction [30], such as play [22], as initial measures to promote coexistence between cats and wildlife.