A team of researchers based in the English Department of the Hong Kong Polytechnic University is compiling the Hong Kong Corpus of Spoken English (HKCSE). The HKCSE comprises three 50-hour sub-corpora: conversations, spoken academic discourses, and spoken business discourses. In this paper we describe some of our findings to date and discuss the learning and teaching implications of the HKCSE in general. The use of tag questions by the two sets of speakers is described. Differences between the two groups of speakers have been found in the frequency of use, syntactic structure of the tag questions use, and the functions they are used to perform. The native speakers (NS) and non-native speakers (NNS) usage of three discourse markers, actually, really, and well, is examined and shown to differ in terms of the discoursal functions the two groups of speakers use them for. The use of vague language by the NS and NNS is compared and it is shown that NNS employ vague language in very similar ways to the NS. There are, however, occasions, when NNS use vague language inappropriately, typically when conveying vagueness by scalar implicature, which may result in miscommunication. The use of inexplicitness has also been examined. The level of inexplicitness exhibited in the conversations is generally lower for NS than for NNS. Finally, the above findings, and the very existence of such a corpus, are shown to have important implications for English language learning and teaching.