Abstract In recent years, the cognitive function of the elderly people has gradually become the focus of psychology, many scholars have analyzed the cognitive function in elderly people from many aspects, but most of the studies were cross-sectional studies, there was little from the perspective of longitudinal research. As we all know, the cognitive function of the elderly was a dynamic process, whether the dynamic process of cognitive function in the elderly was consistent with the results of the cross-sectional studies was unclear. In this study, we aimed to investigate the trend of cognitive function in the elderly people and the differences between individuals. In addition, we further examined what factors would lead to the differences between individuals. The data were derived from the CLHLS, in this study, we used the 4 waves of the data (2002, 2005, 2008, 2011) to investigate the trend of cognitive function in the elderly people and the cognitive function of the subjects were measured using the MMSE scale. In this study, a total of 1834 subjects were included after deletion of the missing values. In addition, the method we used in this study was the latent growth curve model. The latent growth curve model was a method which often used for longitudinal research, it can not only get the overall trend of cognitive function of the elderly, but also get the individual differences of cognitive function. In order to realize the trend of cognitive function of the elderly, we constructed an unconditional latent growth curve model. In addition, based on the result of unconditional latent growth curve model, we further constructed a conditional latent growth curve model to examine what factors would lead to the differences between individuals. In this study, we added time-invariant (gender, years of schooling, smoking and drinking) and time-varying (activities of daily living) covariates to the conditional latent growth curve model to research the influences on cognitive function. Firstly, the results from unconditional latent growth curve model showed that the declining trend of cognitive function in the elderly people was non-linear. That is to say, as time going on, the cognitive function in the elderly people was getting worse and during the survey period, the rate of decline was also different. Secondly, we explored the conditional latent growth curve model. For intercepts, we found that the activities of daily living was negatively predicted cognitive function; Years of schooling predicted cognitive function positively; Gender and drinking predicted cognitive function negatively; smoking has no significant effect on cognitive function. For slope, we found that years of schooling and drinking would positively predict the rate of change in cognitive function of the elderly people. In conclusion, the declining trend of cognitive function in the elderly was non-linear, the higher activities of daily living and the years of schooling were the protective factors of cognitive function in the elderly people. The initial cognitive function of women was significantly lower than that of male, and the cognitive function of non-drinkers was significantly lower than that of drinkers.