Recent research assessing spatial judgements about multisensory stimuli suggests that humans integrate bisensory inputs in a statistically optimal manner, weighting each input by its normalised reciprocal variance. Is integration similarly optimal when humans judge the temporal properties of bimodal stimuli? Twenty four participants performed temporal order judgements (TOJs) about two spatially separated stimuli. Stimuli were auditory, vibrotactile, or both. The temporal profiles of vibrotactile stimuli were manipulated, producing three levels of TOJ precision. In bimodal conditions, the asynchrony between the two unimodal stimuli comprising a bimodal stimulus was also manipulated to determine the weight given to vibrotaction. Unimodal data were used to predict bimodal performance on two measures: judgement uncertainty and vibrotactile weight. A model relying exclusively on audition was rejected based on both measures. A second model selecting the best input on each trial did not predict the reduced judgement uncertainty observed in bimodal trials. Only the optimal maximum-likelihood-estimation model predicted both judgement uncertainties and weights, extending its validity to TOJs. TOJ tasks investigate an important goal of sensory processing: Event sequencing. We discuss implications for modelling this process.