ABSTRACT
The purpose of this study was to identify efficient sampling methods for establishing accurate activity budgets for zoo animals. Seven cotton-top tamarins (Saguinus oedipus) from two zoos were videotaped for multiple 90 min sessions, 3 to 4 days per week for 12 weeks. An activity budget was constructed for each animal using a continuous sampling method to analyze 30 hr of video recording of each animal. These master datasets, reflecting actual behavior, were re-sampled using interval sampling lengths of 0.25, 0.5, 1, 2, 4, 5, 7, 10, 15 and 20 min, and cluster sampling protocols (periodic sessions of continuous sampling) of 10 min x 3, 15 min x 2, 20 min x 1, 15 min x 1 and 10 min x 1 (min x repetitions/90 min sample period) to construct additional activity budgets for each animal. The Canberra similarity index was used to determine the statistical relationship between these activity budgets and those based on the master datasets. As interval length increased, there was a consequent decrease in the accuracy of the associated activity budgets as compared with the master dataset. No cluster sampling protocols yielded activity budgets as accurate as the four shortest interval lengths, but all cluster sampling protocols were more accurate than the three longest interval lengths. All the tested protocols varied in ability to accurately portray animal behavior. Overall, interval sampling provided superior behavioral representations at lower observer input. Results from this study will potentially facilitate the standardization of behavior monitoring protocols at zoos.
