Animal Science Department, Ohio State University, Columbus, OH 43210.
Poult Sci. 2019 Dec 1;98(12):6225-6231. doi: 10.3382/ps/pez475.
Continuous sampling provides the most complete data set for behavioral research; however, it often requires a prohibitive investment of time and labor. The objectives of this study were to validate behavioral observation methods of young broiler chickens using 1) 7 scan sampling intervals (0.5, 1, 3, 5, 10, 15, and 30 min) and 2) an automated tracking software program (EthoVision XT 14) compared to continuous behavioral observation, considered the gold standard for behavior observation. Ten 19-day-old Ross 708 broiler cockerels were included in this study. All behavior was video recorded over an 8-h period, and data were collected using a continuous sampling methodology. The same video files were utilized for analysis for scan sampling and automated tracking software analysis. For both analyses, the following criteria were used to identify which method accurately reflected the true duration and frequency for each behavior, as determined by continuous observation: R2 ≥ 0.9, slope was not different from 1 (P > 0.05), and intercept was not different from 0 (P > 0.05). Active, eating, drinking, and maintenance behaviors were accurately estimated with 0.5-min scan sample intervals. Active, inactive, eating, and maintenance behaviors were accurately estimated with 1-min scan sample intervals. Inactive behavior was accurately estimated with 5-min scan sample intervals. The remainder of sampling intervals examined did not provide accurate estimates, and no scan sampling interval accurately estimated the number of behavior bouts. The automated tracking software was able to accurately detect true duration of inactive behavior but was unable to accurately detect activity. The results of this study suggest that high-frequency behaviors can be accurately observed with instantaneous scan sampling up to 1-min intervals. Automated tracking software can accurately identify inactivity in young broiler chickens, but further behavior identification will require refinement.
连续采样为行为研究提供了最完整的数据集；然而，它通常需要大量的时间和劳动力投入。本研究的目的是使用 1）7 个扫描采样间隔（0.5、1、3、5、10、15 和 30 分钟）和 2）自动化跟踪软件程序（EthoVision XT 14）验证小鸡的行为观察方法，与连续行为观察相比，后者被认为是行为观察的黄金标准。本研究纳入了 10 只 19 日龄的罗斯 708 肉鸡公鸡。在 8 小时的时间内，所有行为都进行了视频记录，并使用连续采样方法收集数据。扫描采样和自动化跟踪软件分析都使用相同的视频文件。对于这两种分析，使用以下标准来确定哪种方法准确反映了通过连续观察确定的每种行为的真实持续时间和频率：R2≥0.9，斜率与 1 不同（P>0.05），截距与 0 不同（P>0.05）。使用 0.5 分钟的扫描样本间隔可以准确估计活动、进食、饮水和维持行为。使用 1 分钟的扫描样本间隔可以准确估计活动、不活动、进食和维持行为。使用 5 分钟的扫描样本间隔可以准确估计不活动行为。检查的其余采样间隔不能提供准确的估计，没有扫描采样间隔可以准确估计行为发作的次数。自动化跟踪软件能够准确检测到不活动行为的真实持续时间，但无法准确检测到活动。本研究结果表明，高达 1 分钟的间隔内，可以使用瞬时扫描采样准确观察高频行为。自动化跟踪软件可以准确识别小鸡的不活动，但需要进一步的行为识别。
