The family resemblance approach to nature of science is receiving increasing attention by science educators since its inception about a decade ago. Many scholars of science education have contributed and continue to contribute to it not only theoretically but also by applying it empirically to a wide range of areas such as curriculum and textbook analyses, pre-service teacher training, undergraduate teaching and, STEM education. This article aims to develop the family resemblance approach further. We do this in several ways. First, we clarify its foundations in a way to reveal that it provides not only a domain-specific, but at the same time a domain-general conceptualization of nature of science. Second, we expand the structure of science as a social institution by adding a new category to it, i.e., the reward system, and justify it. Third, we show that two of the most common elements of the category "practices," namely, observation and experimentation, display the character of family resemblance. Then, we explore this for methods and values in science. Finally, we discuss the possibility of a rapprochement between the family resemblance approach and the consensus view.