This book reports on a two-part study: the validation of a test of spoken English for Norwegian secondary school pupils and the corpus-based investigation of the role played by 'smallwords', such as 'well', 'sort of', and 'you know', in bringing about fluency. The first study builds on the Messickian six central aspects of construct validity to produce a practical framework for test validation. It proceeds to use this framework to identify potential sources of invalidity in the test being examined, and concludes that a principal flaw lies in quality of the band-scale descriptors used in assessing test performance, particularly those relating to 'fluency', which was too vaguely described, and which took little regard to the particular language which helps us to keep going, within and across speaking turns, the smallwords. The second study sets about to explore the concept of fluency, and to expose the extent to which it is acknowledged in the literature to be associated with smallwords, albeit under other names. A study of corpora of the transcripts of test takers judged to be at low or high fluency levels, as well as of young native speakers taking the same test, reveals a clear correlation between smallwords use and fluency measured in mechanical ways. In order to investigate how smallwords were used, a framework, based on relevance theory, is drawn up and used to compare the communicative signals sent through smallwords by the learners with those sent by the native speakers. In conclusion, the findings from the corpus study are drawn on to propose new elements to include in descriptors of fluency, and the implications of the study for classroom practices are discussed. The book should, thus, be of interest to those concerned with the design and validation of tests of the spoken language, as well as those interested in what goes into spoken communication and how to help learners acquire fluency.