'Institutional differentiation' is a prominent feature of the South African higher education system. It is used as a lever to ensure diversity within higher education and to ensure that the system caters for the needs of a diverse student body. However, increasing requirements for all universities to do research and to be rated in similar terms according to a single set of research-related criteria is slowly eroding the basis of differentiation. This research study attempts to understand research and publication milieus and cultures in a differentiated university system that is currently categorised as traditional university, comprehensive university and university of technology. In the study I excluded the comprehensive university. I particularly wanted to explore a university and a university of technology as research and publication milieus, because of the strong distinction between universities and technikons that existed before the advent of democracy in 1994. I used a multiple case study design and I present two case studies to show the relation between an institution's research policy trajectory and the types of researchers who contribute to the research publication count of that university. The research study shows that the traditional university has a well-established research culture moving from research-led to research intensive while the university of technology has an emerging research culture. The study found a different range of academics contributing to the publication count at each type of university. Even though those who publish at both universities are motivated differently they had all been enculturated into a strong research culture, which they acquired at a traditional university. A second finding is that those academics who publish have learnt the 'rules of the academic game', either by informal role modelling or by formal mentoring where senior research active academics make the implicit codes explicit. The third finding is that not having a PhD is a major barrier to career advancement even though publication is not determined by having a PhD. The last finding is that the establishment of a research culture takes place over a long period of time and is not grown overnight. The findings raise questions about the extent to which it can be pre-supposed that all three university types can be measured using the same research performance criteria.

This thesis seeks to demonstrate that the new facilities of Web of Science (WoS) online can be used in new ways to enhance understanding of scholarly communication. It investigates four aspects of scholarly communication: characteristics of highly cited articles, citation levels of collaborative articles, citation levels of multi-disciplinary articles, and patterns of annual citation of highly cited articles. For the first two topics it investigates the WoS category of ‘Information Science & Library Science’ (IS&LS), whereas for the other topics it compares diverse WoS categories in science and social science. Although its main data source is WoS, its investigation of disciplinarity also uses Scopus. The thesis finds: (a) Highly cited IS&LS articles tend to be multidisciplinary and cited late, but are not necessarily first-authored by influential IS&LS researchers, (b) Amongst un-cite IS&LS articles the proportion of collaborative articles has remained almost constant over the past three decades whereas for higher cited articles it has grown steadily with time, (C) In social science subjects the level of citation of multi-disciplinary research are generally similar to that of mono-disciplinary research, whereas in science the citations levels for multi-disciplinary research are substantially lower than that of mono-disciplinary research, and (d) In both science and social science many very highly cited articles continue to be heavily cited more than twenty years after publication. This thesis also introduces and uses an indicator for measuring the extent of collaboration called ‘average partner scores’ and indicates a way in which the subject categories of WoS can be investigated without requiring a licence for the WoS database. Finally, it identifies and addresses some of the technical problems of using WoS online to investigate scholarly communication.