Form-Focused Instruction (FFI) has been extensively studied, yet past research often yields contradictory results. Overly simplistic examination of grammatical features, along with inadequate exploration of EFL learners at different English proficiency levels, has limited understanding of how and when to use different forms of FFI. Due to a need for further research, the present study was designed to holistically examine input and output-based forms of FFI with various grammatical features, along with Korean EFL learners at different proficiency levels. Eighteen experimental studies with Korean EFL learners were selected for metanalysis and research synthesis. The effects of type of instruction (input vs. output) were analyzed along with complexity of a target feature and learner proficiency in English. Results suggested that input-based instruction is more effective with grammatical features that involve a single phrase (e.g., basic morphology and vocabulary), while output-based instruction is more effective with more complex phrases and clauses with a syntactic component (e.g., relative clauses and conditionals). Input may activate form-meaning mapping, which is more useful for the acquisition of basic morphology or vocabulary, whereas output may help learners order constituents, thereby aiding in the acquisition of syntactic phrases and clauses. Proficiency also appears to influence the efficacy of FFI. When proficiency of participants was low and complexity of a grammatical feature was high, the influence of input tended to be stronger. When proficiency was intermediate and complexity of a grammatical feature was low, output-based instruction tended to result in larger effect sizes.