Abstract: Objective To choose a better version of Rivermead Behavioral Memory Test (RBMT) to assess memory function of patients with mild traumatic brain injury (TBI). Methods From April, 2015 to Febrary, 2017, 40 mild TBI patients and 40 healthy people were recruited as TBI group and control group respectively. Both groups completed the Chinese version of RBMT-II first, and 24 hours to 48 hours later, completed the Chinese version of RBMT-III. The raw score of each test and the number of perfect scores and floor performance were scored and compared. Results Compared with the control group, TBI group got lower scores in six subtests of RBMT-II (F>2.131, P<0.05) and twelve subtests of RBMT-III (F>2.035, P<0.05). Administration of the RBMT-III resulted in less participants performing at or near individual subtest's ceiling than RBMT-II, mainly in the picture recognition, face recognition, the line instant memories, the line delay memories, letters delayed recall and orientation date (Z>2.117, P<0.05). Also administration of the RBMT-III resulted in less floor performance than those of RBMT-II, mainly in remembering the name and the appointment (Z>2.138, P<0.05). Conclusion RBMT-III has substantial improvement over the original RBMT-II, as it reduces the problem of ceiling and floor performance and the number of misclassifications.

Key words: mild traumatic brain injury, Rivermead Behavioral Memory Test, ceiling effect, floor effect