The objective of our study was to assess the correlation of routine neuropsychological test results in elderly patients referred to a gerontopsychiatric ward. MMSEs, CTs and SKTs were performed in 94 patients (age: median = 74 years, range = 54-89 years; 64 f, 30 m) with mild to moderate dementia and evaluated retrospectively. Pairwise Spearman rank correlation, multiple regression and response surface analysis were used to assess relations between test results. The scores of all three tests used were reciprocally correlated (P < 0.05) yielding the following pairwise R-values: SKT versus MMSE: -0.77, SKT versus CT: 0.69, MMSE versus Clock test: -0.61. Multiple regression analysis showed a maximum correlation of 0.87 and marked standardised beta values, if SKT was chosen as dependent variable. Test scores could be well fitted to both symmetric linear (SKT = a + b x CT + c x MMSE; R(2) = 0.67, P < 0.01) and non-linear (SKT = a + b x CT + c x MMSE + d x CT(2) + e x MMSE(2) + f x CT x MMSE; R(2) = 0.67, P < 0.01) response surfaces. In conclusion, test scores of SKT, CT or MMSE in patients with dementia showed a marked correlation. In clinical practice the may be considered as almost interchangeable test options.