Aim: To observe the damage of the insight in dementia patients and to evaluate the effect of memory complaints in the clinical diagnosis of dementia. Methods: Fifty-four dementia patients and 50 controls were included in the study. All of the subjects and their informants were examined with Informant Questionnaire of Cognitive Decline in the Elderly (IQCODE) separately to observe the difference in assessment of cognitive impairment. Two additional questions were asked: 1 Did the subjects have hypomnesia? 2 Who found the hypomnesia first? Information-Memory-Concentration Test (IMCT) was tested as an objective parameter reflecting cognitive function. The Center of epidemiological Survey-Depression Scale (CES-DS) was tested to observe the relations between the self-assessment and informants assessment. Results: There were significant differences in IQCODE of informants (3.89 ± 0.44 vs. 3.30 ± 0.26) and IMCT scores (15.30 ± 8.21 vs. 28.22 ± 5.81) between the two groups, while there were no significant differences in self-assessment (3.74 ± 0.56 vs. 3.63 ± 0.38) and depression scores (13.34 ± 7.13 vs. 15.10 ± 8.22) between the two groups. The correlation between informant assessment and IMCT was significant (r = 0.71, P < 0.05) as well as between self-assessment and depression scores in the two groups (r = 0.68, P < 0.05). In the dementia group there were more informants that recognized the hypomnesia in patients than the patients themselves (X 2 = 15.81, P < 0.01). While in the control group there were fewer informants that recognized the hypomnesia than the patients themselves (X 2 = 11.52, P < 0.01). Informants were a majority of persons who firstly found the hypomnesia in the dementia group X 2 = 6.52(0.01 < P < 0.05). Contrarily, the subjects were a majority of persons who firstly found the hypomnesia in the control group (X 2 = 3.23, P > 0.05). Conclusion: As the insight of dementia patients is impaired, there is no significant correlation between the self-assessment and objective assessment. The self-assessment has a low reliability in clinical diagnosis of dementia. The qualified informants can offer the valid information for clinical diagnosis of dementia.