Working memory (WM) enables the manipulation of information necessary for goal-directed behavior. What limits WM capacity? Interference between representations has been identified as one factor, degrading WM representations, rendering them inaccessible. This study focuses on interference arising from “code occupation”. Code occupation denotes a process whereby a feature bound in one representation is temporarily occupied and therefore less available for binding to new representations. Accordingly, performance should worsen when items share features compared to when they do not, yielding partial overlap costs. Such costs are consistently observed for procedural WM representations, specifically representations of planned but not yet executed actions. Declarative WM, namely memory for items, typically does not exhibit partial overlap costs. This study investigates whether declarative WM is affected by code occupation. To this aim, we designed a closely matched paradigm to measure partial overlap costs in procedural (Exp. 1) versus declarative WM (Exps. 2 and 3). In Experiment 1 (N=60), participants performed an action while holding another action plan in procedural WM. In Experiments 2 and 3 (Ns=40), participants memorized an item and performed a two-alternative forced-choice recognition test on it while holding another item in declarative WM. Across all experiments, we manipulated features overlap of the to-be-maintained action plans or items. We observed partial overlap costs in both procedural and declarative WM, indicating that code occupation also limits declarative WM capacity. This result speaks for shared mechanisms across WM systems.