The aim of this study was investigated a maze learning in stroke individuals. Forty participants assigned into two groups: experimental (stroke participants; n = 20) and control (neurologically healthy participants; n = 20). The study involved an acquisition phase, a transfer test, and a short-and longterm retention tests. The task consisted in complete a maze, with paper and pen, in the shortest time possible. The dependent variables were execution time and error. Data were analyzed with an Anova- two way with Repeated Measures for these variables. Results showed learning for both groups, but with the experimental group having worse performance compared to control group mainly related error. It was also seen the impact promoted in the task has impaired both groups in the transfer test performance.