Inferring emotions from Head Movement (HM) and Eye Movement (EM) data in 360° Virtual Reality (VR) can enable a low-cost means of improving users’ Quality of Experience. Correlations have been shown between retrospective emotions and HM, as well as EM when tested with static 360° images. In this early work, we investigate the relationship between momentary emotion self-reports and HM/EM in HMD-based 360° VR video watching. We draw on HM/EM data from a controlled study (N=32) where participants watched eight 1-minute 360° emotion-inducing video clips, and annotated their valence and arousal levels continuously in real-time. We analyzed HM/EM features across fine-grained emotion labels from video segments with varying lengths (5-60s), and found significant correlations between HM rotation data, as well as some EM features, with valence and arousal ratings. We show that fine-grained emotion labels provide greater insight into how HM/EM relate to emotions during HMD-based 360° VR video watching.