This study investigates several aspects of Chinese shi tag questions, including their distributional properties, the sequential structures of those questions as adjacency pairs, and their functions in conversational contexts. Unlike previous studies using invented data to analyze tag questions, this study focuses on shi tag questions in naturally occurring conversation. In particular, the questions, shiba, shibushi, shima and bushima are examined for their respective functions. Analysis of the data shows that the functions of shi tag questions can be classified into three major categories: 1) certaining the truth or falsity of previous information delivered by the speaker him/herself; 2) checking another interlocutor’s previous information by asking for it again; and 3) performing social, interactional functions. Chinese shi tag question utterances seem to function in conversation by inducing the interlocutor’s collaboration, understanding, and preferred answers within their common speech actions. These findings suggest linkages between interaction and discourse functions for shi tag questions and the study may offer support from Chinese to the hypothesis that grammaticalization is a result of repeated daily, routinized use of language.