Based on grabing and structural collection of 58386 pieces of news from June 2012 to April 2016 and their emotional feedback in “social new” section of sina, by empirical and quantative method, this paper examined and analysized emotional bias and its features on internet. Not only contents in media have emotion, but also media have emotion. In the samples of social news on sina investigated, there is a significant "angry" bias, and its effects and characteristics are as follows: first, the scale and proportion of anger are bigger than other emotions; second, the distribution of various emotions is more biased to and concentrated on anger when anger is dominant mood; third, there is a more closely positive correlation between the emotional intensity and the anger than other emotions; fourth, there is a more closely positive correlation between the spread of news and anger than other emotions. Emotional bias in media is also helpful to the theoretical and practical problems such as bias in the formation of “spiral of silence”, effect and pattern of emotional setting and the mechanism and strategies of emotional communication.