Behaviour 2019
Timing is Everything: Exploring the Optimal Timing for Sampling Behavior Data
Colin M Lynch, Ioulia Bespalova, Jennifer H Fewell. Arizona State University, Tempe, AZ, United States

Ethologists often encounter time constraints when attempting to categorize animal behavior. Even in cases when the entirety of an animal’s behavioral repertoire has been video recorded, it might not be possible or necessary to analyze all the available data. But, how do you choose what segments to sample, to ensure that the sampling is representative of the animal’s overall behavioral paradigm? In this study, we determine the task preference of several Pogonomyrmex californicus ants by analyzing three to six hour long videos. We then simulate an ethologist sampling from the video by randomly drawing from this dataset using different scanning protocols. In these protocols, a sample can either be a singular block, or segmented and spread throughout the dataset. We find that increasing the number of small segments of a sample (instead of having fewer, large blocks) usually increases the probability that the sample is indistinguishable from the entire dataset. We also examine characteristics of the dataset that, in some cases, allow samples with fewer segments to be the most similar to the total dataset.