Previous studies have demonstrated that humans use a variety of facial cues, including gaze direction and nose angle, to perceive the direction in which a crowd is looking. For example, in a study using simple line-drawing faces, Sweeny et al. (2014) argued that the visual system pools information from nose angle and pupil rotations at the group level to determine the focus of attention of a crowd. However, there is currently no consensus as to the relative importance of eyes and nose cues in determining the focus of attention. Furthermore, the relative importance of these cues may differ when perceiving individual faces compared to crowds. The purpose of the first experiment was to determine whether the ensemble perception of the focus of attention of based on more than one face.For investigating the process of perception, Experiment 2 used the eye tracker to explore differences in how the focus of attention is derived when looking at crowd vs individual people. In experiment 1, either an individual face or crowd of faces (two, three or four faces) were presented on a 14-inch Lenovo computer by E-prime 2.0. Ninety-six faces with various pupil rotations and head orientations were constructed by Autodesk Maya 2014 and Photoshop 5.0. Thirty undergraduate students were required to complete an averaging orientation judgement task in which they wereinstructedtoestimate where the individual’s face or crowd’s faces were looking. Experiment 2 was 2 (facial cues: eyes and nose) × 4 (number of faces) within-subject design. Thirty-four participants` eye movements were recorded with an SMI RED eyetracker(22-inch screen, sampling rate=250Hz, resolution=1024 ×768 pixels, refresh rate=60Hz, participants - and - screen distance=57-60cm). The results of experiment 1 showed that the accuracy of using multiple faces of the crowd to estimate the direction of crowd’s attention was significantly higher than that of using an individual’s face (p＜0.05＝, which indicated that observers integrated at least two faces to perceive crowd attention. The results of experiment 2 suggested that the way in which we estimate an individual’s attentional focus is distinct from how we estimate the focus of attention of a crowd. (1) For both individual and crowd of faces, the entry time of noses was significant lower than that of eyes, p＜0.05, (2) The first fixation duration of the noses on individual faces was longer than that of eyes, p＜0.01, whilefor the crowd of faces, there is no significant difference between the first fixation duration on the noses and eyes. (3) The probability of first fixation located on noses was higher than that of eyes in individual faces condition, p＜0.05, yet in two faces or four faces conditions, the probability of first fixation located on eyes was higher than that of noses, p＜0.01. (4) For individual faces, total fixation duration of noses was longer than that of eyes, p＜0.01, whereas for crowd faces, there is no difference between noses and eyes on total fixation duration. When determining the focus of individual’s attention through facial cues, nose is more important than eyes. While under crowd condition, both nose and eyes are equally important. The results indicated that the visual system pools the gazes of faces into an ensemble code based on both eyes and nose, which allows humans to rapidly and efficiently perceive where a crowd is looking.