Existing research shows that a sad mood hinders emotion recognition. More generally, it has been shown that mood affects information processing. A happy mood facilitates global processing and a sad mood boosts local processing. Global processing has been described as the Gestalt-like integration of details; local processing is understood as the detailed processing of the parts. The present study investigated how mood affects the use of information processing styles in an emotion recognition task. Thirty-three participants were primed with happy or sad moods in a within-subjects design. They performed an emotion recognition task during which eye movements were registered. Eye movements served to provide information about participants' global or local information processing style. Our results suggest that when participants were in a happy mood, they processed information more globally compared to when they were in a sad mood. However, global processing was only positively and local processing only negatively related to emotion recognition when participants were in a sad mood. When they were in a happy mood, processing style was not related to emotion recognition performance. Our findings clarify the mechanism that underlies accurate emotion recognition, which is important when one is aiming to improve this ability (i.e., via training).