(A) Experimental protocol to test non-cued spatial discrimination by delayed non-matching-to-place (DNMP) using an 8-arm radial arm maze. Food restricted animals were placed in the start (S) arm and the froot loop reward (R) placed in the goal arm. (B) Learning curves for each genotype over a 5-day experiment. Animals were randomly assigned two trials in the two-arm separation and two trials in the four-arm separation each day and the results averaged and expressed as percent correct choice. The results were analyzed by two-way ANOVA and showed an effect of genotype in the two-arm separation (KO:WT, p=0.004, DG-KO:HET, p=0.013). In the four-arm separation the KO:WT comparison was significant, (p=0.012); the DG-KO:HET comparison was not significant (p=0.095). (C) AKAP7 global KO and DG-KO animals discriminate poorly compared to their respective controls in the more difficult two-arm separation test. The performance on days 4–5 in the two-arm separation test was pooled and analyzed by two-tailed, unpaired T-test (WT:KO, p=0.023, Cohen's d = 1.04; HET:DG-KO, p=0.025, Cohen's d = 1.01). In the four-arm separation trials the WT:KO and HET:DG-KO comparisons did not show significant differences (p=0.565 and p=0.346 respectively). WT and HET groups in both the two-arm and four-arm trials performed significantly better than 50% at days 4–5 (p<0.05), whereas the KO and DG-KO did not. KO (n = 11), WT (n = 11), HET (n = 14), and DG-KO (n = 9). *p<0.05, **p<0.01.