Humans can intentionally forget previously-learned declarative information such as words: Memory for to-be-remembered (TBR) words is typically better than for to-be-forgotten (TBF) ones (directed forgetting effect). The role of intention in the learning and retrieval of procedural bindings is, however, less clear. Here, we combined item-method directed forgetting and item-specific S-R priming to investigate whether people have intentional control over the formation and/or retrieval of stimulus-response (S-R) associations: By categorizing stimuli in a learning phase participants formed S-R associations. A memory cue following participants’ responses indicated to remember or forget the stimulus for a later memory test. In the following test phase, participants responded to the same stimuli, but the required response item-specifically repeated or switched. In four experiments, reaction time differences between item-specific repetitions and switches (item-specific S-R effect) did not differ between TBR and TBF stimuli. This was the case when S-R associations already existed before the memory instruction was given (Exp. 1), when stimuli, responses, and memory cues were paired multiple times (Exp. 2), when forming new S-R associations (Exp. 3), and when categorizing word stimuli rather than images (Exp. 4). Thus, intentionally up- and/or downregulating the memory strength of a stimulus representation does not necessarily go hand in hand with the strengthening of the associated response. These results can be explained by two accounts: either top-down control can selectively operate on different levels of representations (and does not incidentally affect an entire event file) or that separate memory systems store procedural and declarative content.