The consolidation of long-term memory is thought to critically rely on sleep. However, first evidence from a study in Drosophila suggests that hunger, as another brain state, can benefit memory consolidation as well. Here, we report two human (within-subjects crossover) experiments examining the effects of fasting (versus satiated conditions) during a 10-hour post-encoding consolidation period on subsequent recall of declarative and procedural memories in healthy men. In Experiment 1, participants (n = 16), after an 18.5-hour fasting period, encoded 3 memory tasks (word paired associates, a visual version of the Deese-Roediger-McDermott task, finger tapping) and subsequently either continued to fast or received standardized meals. Recall was tested 48 h later in a satiated state. Experiment 2 (n = 16 participants) differed from Experiment 1 in that a What-Where-When episodic memory task replaced the Deese-Roediger-McDermott task and recall was tested only 24 h later in a fasted state. Compared with the satiated state, fasting enhanced cued recall of word paired associates (more correct and faster responses) and item recognition in the What-Where-When task. By contrast, fasting impaired recall of episodic context memory, i.e., spatial context in the Deese-Roediger-McDermott task, and temporal-spatial context in the What-Where-When task. Procedural memory (finger tapping) remained unaffected. The pattern suggests a differential effect of fasting selectively promoting consolidation of semantic-like representations in cortical networks whereas hippocampal representations of episodic context are weakened. We speculate that hunger strengthens cortical representations by suppressing hippocampal interference during wake consolidation. Yet, the underlying mechanism remains to be clarified.