Although abstraction is widely understood to be one of the primary components of computational thinking, the roots of abstraction may be traced back to different fields. Hence, the meaning of abstraction in the context of computational thinking is often confounded, as researchers interpret abstraction through diverse lenses. To disentangle these conceptual threads and gain insight into the operationalisation of abstraction, a systematic review of 96 empirical studies was undertaken. Analysis revealed that identifying features of entities, extracting relevant features, discovering patterns, creating rules and assembling the parts together were the core actions of abstraction. With the primary aim of simplifying practical procedures, abstraction was operationalised as the sophistication of a program, the matching of patterns, the creation of alternative representations, the transfer of solutions, the measurement of a learner's activity and reading program codes. There is an obvious need for researchers to align the conceptual meanings they have established of abstraction with the practical facts of operationalisation. The need to empirically validate emerging models and the implications for future research are discussed.