Many emerging smart applications and services employ Web technology, and users nowadays surf the Web from any device via any kind of access network. Typically, high page latencies trigger users to abort ongoing transfers, resulting in the abrupt terminations of the TCP connections. This paper presents a systematic study of the termination process of the TCP connections and identifies the reasons behind the observed sequences of termination flags. Monitoring and classification of the termination behavior of the TCP connections can provide indications about the user-perceived performance of Web transfers. From the results, it is observed that the TCP termination behavior is heavily-dependent on the client-side application. Therefore, a set of criteria is required to identify the abortions made by the user.