To create and evaluate an inexpensive novel model for training in retroperitoneal laparoscopic dismembered pyeloplasty. The retroperitoneal cavity was simulated by five boards fixed together by detachable hinges; the extent of operating difficulty could be adjusted by changing the angle of inclination. The middle of a plasticine model kidney was imbedded into a metal clip, and a carp swim bladder was fitted to the kidney to model the dilated pelvis. The "ureter" was a 10-cm long porcine ureter. Semicircular-shaped simulator, laparoscopic camera system, scissors, needle holder, and surgical forceps specific for training were used. Ten rounds of intensive training of the "ureteropelvic" anastomosis procedure was provided for five laparoscopic surgeons. Each trainee's mean time of achieving dismembered pyeloplasty in three patients was recorded; the corresponding index in the last three ureteropelvic junction obstruction operations performed by each person before intensive training was retrospectively revised. Subjective evaluation was also performed, and a general self-efficacy (GSE) scale was applied to evaluate the self-confidence of trainees. This model reproduced the technical complexity of laparoscopic dismembered pyeloplasty. The operative time significantly declined from 41.84 +/- 3.00 minutes to 25.04 +/- 2.37 minutes after intensive training. GSE scores after training were also higher than before (27.60 +/- 1.14 vs 22.20 +/- 1.79). Correlation between operative time and GSE was significant. After training, all participants were more skillful in performing ureteropelvic anastomosis. This is a small study with a middle-term follow-up. The operation also cannot be simulated completely because the ureter is separate from the renal pelvis in the beginning. This model demonstrates satisfying fidelity and practicality; it could easily be built and reused, and could result in lower costs than those incurred using the traditional model.