Many studies demonstrated a higher accuracy in perception and action when using more than one sense. The maximum-likelihood estimation (MLE) model offers a recent approach on how perceptual information is integrated across different sensory modalities suggesting statistically optimal integration. The purpose of the present study was to investigate how visual and proprioceptive movement information is integrated for the perception of trajectory geometry. To test this, participants sat in front of an apparatus that moved a handle along a horizontal plane. Participants had to decide whether two consecutive trajectories formed an acute or an obtuse movement path. Judgments had to be based on information from a single modality alone, i.e., vision or proprioception, or on the combined information of both modalities. We estimated both the bias and variance for each single modality condition and predicted these parameters for the bimodal condition using the MLE model. Consistent with previous findings, variability decreased for perceptual judgments about trajectory geometry based on combined visual-proprioceptive information. Furthermore, the observed bimodal data corresponded well to the predicted parameters. Our results suggest that visual and proprioceptive movement information for the perception of trajectory geometry is integrated in a statistically optimal manner.