Different output from Nonparametric Tests Chi-Square vs. Crosstabs Chi-Square

Problem

I would like to test the significance of the association between two categorical variables, GROUP and RESPONSE, with a chi-square test. I entered GROUP and RESPONSE as the Row and Column variables in the CROSSTABS procedure (Analyze->Descriptive Statistics->Crosstabs) and requested the Chi-square statistic in the Statistics dialog for that procedure. I found another chi-square test in the NPAR TESTS procedure (Analyze->Nonparametric Tests->Chi-Square) and entered GROUP and RESPONSE in the 'Test Variable List'. The chi-square results from the two procedures do not agree. In fact, NPAR TESTS seems to report a chi-square for each of the GROUP and RESPONSE variables. Also, CROSSTABS reports that there are some cells with expected frequencies less than 5, whereas the Chi-Square results in NPAR TESTS report that there are 0 cells with expected frequencies less than 5. What is the reason for the discrepancy in chi-square results between these procedures?