Major depression (MD) is associated with increased cardiovascular risk. Although alterations in autonomic regulation have been proposed as one potential pathophysiological mechanism to explain this comorbidity, studies using standard HRV features in depressed subjects have been inconclusive. In this study, 48 patients with MD and 48 healthy controls (HC) were randomly assigned to an audio-visual task with two different versions: one emotionally neutral (N) and the other emotionally arousing (E). ECG signal (lead II) was collected at 250 Hz, and point process nonlinear analysis of heartbeat dynamics was performed to obtain instantaneous features from standard time-domain analysis, as well as spectral (LF, HF, LF/HF) and bispectral (LL, LH, and HH) analysis. Mean values of all features were computed over the 30s segment of the emotional elicitation session. Only bispectral parameters LH and HH were significantly different between patients and HC (p<;0.02). Our results suggest that time-varying nonlinear dynamics of parasympathetic activity are significantly reduced in MD compared to HC in response to emotional elicitation. We conclude that instantaneous bispectral analysis could be a promising tool for assessment of autonomic modulation in MD.