Abstract: Back-calculation, which uses a set of measurements from the bony structures in the age determination and its current body length at one time to infer its length at an earlier time or times, is a very important tool in fisheries research and management. Because a lot of bony structures can be used in age determination, and the relationship between the chosen structure and the body length of fish also can be described with various linear or non-linear equations, several back-calculation models have been proposed. In order to expand confidence in estimation of growth parameters of a particular fish species, it is necessary to obtain a comparison of back-calculated lengths from different structures and models.The present study dealt with the suitable calcified tissues and methods used for growth back-calculation of roach Rutilus rutilus (Linnaeus, 1758) from Ulungur Lake, the second largest lake of Xinjiang Uigur Autonomous Region, one of the tenth freshwater lakes in China. Two calcified tissues (scales and lapilli) and four common back-calculation methods (Dahl-Lea, Regression, Fraser-Lee and BPH) were utilized for growth back-calculation. The reliability was evaluated by comparing the back-calculated lengths at ages with observed lengths at the annuli formation. It was shown that the back-calculated lengths obtained by the lapilli were more reliable than those obtained by scales, with smaller differences from observed lengths. So, it was demonstrated that the lapilli were the better material both for age determination and back-calculation of roach. For scales, the Dahl-Lea method provided the most unreliable estimates of fish lengths at previous ages, with the largest differences from the observed length (?26.0%), it also had significant differences from the other three methods. The differences between the back-calculated lengths from the Regression, Fraser-Lee and BPH method and the observed lengths were very close, which was ?6.7%, ?7.0% and ?7.0% respectively. It indicated that the three methods were equal for the back-calculation of roach. For lapilli, the Dahl-Lea was the best method for the back-calculation of roach with the smallest (?1.9%) difference from the observed lengths. The Regression method was the worst one with the largest difference from the observed lengths (?4.5%). The Fraser-Lee method provided the similar results with the BPH method and the differences from the observed lengths were both between ?1.4% and ?3.2%, averaged ?2.6%. They were secondary for growth back-calculation of roach.