A focus of recent research is to understand the role of our own response goals in the selection of information that will be encoded in episodic memory. For example, if we respond to a target in the presence of distractors, an important aspect under study is whether the distractor and the target share a common response (congruent) or not (incongruent). Some studies have found that congruent objects tend to be grouped together and stored in episodic memory, whereas other studies found that targets in the presence of incongruent distractors are remembered better. Our current research seems to support both views. We used a Tulving-based definition of episodic memory to differentiate memory from episodic and non-episodic traces. In this task, participants first had to classify a blue object as human or animal (target) which appeared in the presence of a green one (distractor) that could belong to the same category as the target (congruent); to the opposite one (incongruent); or to an irrelevant one (neutral). Later they had to report the identity (What), location (Where) and time (When) of both target objects (which had been previously responded to) and distractors (which had been ignored). Episodic memory was inferred when the three scene properties (identity, location and time) were correct. The measure of non-episodic memory consisted of those trials in which the identity was correctly remembered, but not the location or time. Our results show that episodic memory for congruent stimuli is significantly superior to that for incongruent ones. In sharp contrast, non-episodic measures found superior memory for targets in the presence of incongruent distractors. Our results demonstrate that response compatibility affects the encoding of episodic and non-episodic memory traces in different ways.