Objective: The aim of this study was to develop a new and shorter Chinese version of Ryff's psychological well-being scale. Design: Cross-sectional survey. Setting: In recent years there have been several versions of this scale, including 84-item, 54-item and 18-item versions. Researchers in different countries have built on Ryff's version to create their own versions in various languages. However, most versions have not yielded improved confirmatory factor analysis results. Shorter versions of the scale in the past were also less reliable in terms of internal consistency. Method: This study used a sample of 820 middle-aged and older people from Taiwan with an average age of 59.8 years. We used structural equation modelling (SEM) to select items for inclusion. Results: After item selection, a new version of 18 items based on Ryff's 84-item version was established with improved confirmatory factor analysis validity and reliability. The results for this 18-item version indicated that factor loadings were at least 0.60, and the reliability alpha coefficients were 0.92 and 0.60--0.75 for its six subscales. Additionally, a significant correlation with the Geriatric Depression Scale (GDS)-15 and quality of life criteria indicate very acceptable criterion-related validities. Conclusion: The 18-item Chinese version does not include any reversely worded items after the item selection processes, thus reducing method artefacts. It also exhibited very acceptable confirmatory factor analysis results.