In spoken languages, new information is often produced with a longer word duration than given information. We investigate whether signers use the prosodic cue of duration to mark information status. Thirty-two deaf Chinese Sign Language (CSL) signers retold a story after watching a short cartoon clip. The data were glossed by a native CSL signer, with target references coded (e.g., different mentions of ‘bear’, ‘stone’). We examined whether there was any reduction in the references as a function of the information status. The results showed that first CSL signers mostly used nominals (24.0%), classifiers (42.1%), zero anaphora (30.0%), but hardly any pointings (3.9%). The nominals were both used in the first and subsequent mentions whereas classifiers, zero anaphora and pointings were almost always used in the non-first mentions. Furthermore, focusing on nominals, we compared sign durations over five mentions (M1=667.43ms; M2=426.09ms; M3=397.88ms; M4=440.03ms; M5=494.61ms). A regression analysis showed a significant linear and curvilinear relationship, indicating a gradient decrease in sign duration for the first three mentions and climbing up for the fourth and fifth mentions. In conclusion, CSL signers not only use linguistic devices to track references but also vary the duration of nominals to mark information status.