Urban vacant land has been concealed for a long time in the contradictory but unified processes of urban growth and shrinkage. The study on urban vacant land, however, is currently a rather rudimentary and subjective issue. This paper firstly presents the connotation of urban vacant land based on bibliometric analysis. Then, typology, proximate causes and values of urban vacant land are analyzed at parcel, transect, city and national levels. Results show that the study in China lags about half a century compared with North America and Europe. Moreover, geographers has seldom paid attention to this issue. Urban vacant land could be proximately categorized by land cover, land usage and land ownership. It is widespread with a large amount in cities. For example, the probabilities of urban vacant land in the cases of Guangzhou and New York are 8.46%-8.88% and 3.17%-5.08%, respectively. In addition, the average proportion of urban vacant housing land take up 11.48% in the 65 cities in the USA. Generally, urban vacant land shows fragmentation and odd shape. Significant spatial differences exist at parcel, transect, city and national levels. Proximate causes, such as land division, odd shape, demographic decline, de-industrialization, land speculation, disinvestment, and environment issues could result in urban vacant land, which has become the Grey Island in urban social, economic and ecological spaces. However, it could be considered as the potential resources for enhancing urban sustainability. To promote urban renewal and land supply-side structural reform, urban vacant land research framework is finally discussed. Specifically, the logical research themes on urban vacant land consist of unified typology, dynamic process and pattern, underlying driving forces, transformation mechanisms among different types, impacts and multifunction assessment, and sustainable use control. Focus should be given to research on social, institutional, economic and ecological interaction, comparative study in different urbanization processes and regions, as well as multi-scale study.