Introduction: The objective was to evaluate the factor structure and propose a new version of the Rosenberg Self-Esteem Scale with only positive items to overcome the method effect associated with negative items. Method: A version A (positive and negative items) and a version B (only positive items) were considered. A sample of 350 university students was collected for each version. Results: The CFA shows that version A's one-dimensional model (A1) does not present adequate fit indices. It was also found that adding a specific factor for negative items (model A2) and another factor for positive items (model A3) does not improve the fit indices. A two-dimensional model (A4 model) does not improve the fit indices either. Regarding the one-dimensional model of version B, it presents a superior fit compared to the original model (model A1) Discussion and Conclusions: It is concluded that version B, a proposal of only positive items, adequately measures self-esteem since it does not have negative items and agrees with Rosenberg's original approach.