Abstract This study analyzed the alarm calls of Formosan macaques and their functions from Mt. Longevity. Under the natural condition, 137 five-minute scan sampling and 129 completely 20-minute all occurrence sampling were collected to record behaviors and the alarm calls. In addition, 43 stimulus (predator) tests and 87 playback experiments were successfully conducted from September 2002 to March 2003. Incidents that triggered alarm calls included: the confrontation with dogs, threats from travelers with a cane, slingshots, or stones, encounters with other troops of macaques, passing-by motors or mobile cars, airplanes flying above, and some unrecognizable factors. Under the natural condition, most of the alarm calls produced by Formosan macaques were responds toward dogs (63.46%), with the average frequency of 0.78 times/10hrs. In the predator tests, types of stimulus (dog, human with slingshot, m-snake1 and m-snake2), sex/age classes (adult male, adult female, juvenile, and infant), and the position of macaques (0m, <1m, 1m, and ¡Ù2m) had great influence on the behavioral response of macaques (p<0.01). The response scores were from 3 to 0 (move away more than 5 times body length or climbed to tree, move away up to 5 times body length, visual orientation towards the predator and no apparent response). The average response score of macaques to dogs was the highest one, far above threats from a person with a slingshot, m-snake1 and m-snake 2 (p<0.05). When Formosan macaques confront dogs, the vocal frequency (time/individual) to the alarm call had great influence on sex/age classes of macaques (p<0.01), and the average alarm call frequency was highest from adult males. When human with slingshots and m-snake were predators, majority of the adult males and females adopted run-away (69.91¢M), while very few climbed up trees (8.02¢M) to prevent from any harm caused by predators. However, they ran away (31.52%) or climbed up trees (48.47%) in response to dogs in different proportions. The average response score of macaques, from high to low, was from infant, juvenile, adult female and adult male, and the differences were significant (p<0.05). When Formosan macaques encounter these four types of predators, their alarm calls were quite similar in the spectrographs. The six basic vocal characteristics (maximal, median and minimal fundamental frequency, lowest and peak frequency and duration), analyzed by canonical discriminate analysis, indicated that alarm calls of Formosan macaques confront dogs and human with slingshots could be distinguished from snake models. The four types of predators had significant effects at the median and maximal fundamental frequency of the basic vocal characteristics from adult females and juveniles (p<0.01). But there was no difference in the nine basic vocal characteristics of alarm calls from adult males and infants toward four types of predators. The alarm calls of adult females and juveniles toward dogs in the median and maximal fundamental frequency both were significantly lower than those from m-snake1. In the playback experiments, sex/age classes, types of the alarm calls toward stimulus (dogs, travelers with slingshots, m-snake1 and m-snake 2) and the position of macaques had significant effects on their behavioral responses (p<0.01). The average response score of macaques in playbacks, from high to low, was from infant, juvenile, adult female, and the adult male, and the differences were significant (p<0.05). When the alarm call caused by dogs played back, the average response score of macaques was higher than the alarm calls stimulated by m-snake1 and m-snake 2 (P<0.05). In the playback experiments, when macaques at a higher place (¡Ù2m), they often visually orientated towards the predator or no apparent response whit a lowest response score. The average response scores of the adult males and females toward four types of alarm calls (playback) were higher than the control ones (p<0.05).