In recent years, numerous research projects have introduced computational thinking to primary school teachers as an initial step in exposing students to computer science concepts at an early age. One consistent difficulty across projects that provide computational thinking professional development to elementary teachers is teaching abstraction. This position paper makes preliminary recommendations for abstraction pedagogy in elementary education based on an examination of a series of unplugged abstraction examples demonstrated during a summer computational thinking professional development institute. By examining commonalities among abstraction examples, key parts of the process of abstraction pertinent to elementary classrooms were identified: (1) the abstraction process is typically performed in reverse since students in elementary school are typically given abstractions to start with; (2) evaluation of details against a valid set of criteria for an abstraction is part of the filtering step of abstraction; (3) in the absence of a valid set of criteria, pattern recognition can be applied to a set of valid examples to extract characteristics of an abstraction; and (4) abstractions can be supported by details that also happen to be abstractions which triggers additional abstraction steps. Preliminary recommendations for abstraction instruction include having students evaluate examples; engaging students in pattern recognition to extract characteristics of an abstraction; developing student fluency in describing abstractions, their supporting examples, and characteristics; and assessing students by asking not only for examples of abstractions, but for their characteristics as well.