Fig. 4

Stimulus pair retention and transitive inference is intact in Nlgn3^−/y and Nlgn3^R451C mice. Performance accuracy (% correct responses) on the 4 premise pairs during a pseudorandom integration (stage 4, last 5 sessions) and b transitive inference test session. a Relative to A⁺B⁻, B⁺C⁻: p < 0.001, aMD = − 0.205, 95% CI = − 0.254, − 0.156; C⁺D⁻: p < 0.001, aMD = − 0.082, 95% CI = − 0.121, − 0.043). C⁺D⁻ greater than B⁺C⁻ (p < 0.001, aMD = 0.123, 95% CI = 0.082, 0.165. D⁺E⁻ marginally higher than A⁺B⁻ (p = 0.043, aMD = 0.045, 95% CI = 0.001, 0.088. B B⁺C⁻ compared to all other pairs (A⁺B⁻: p = 0.007, aMD = − 0.25, 95% CI = − 0.432, − 0.068; C⁺D⁻: p = 0.002, aMD = − 0.25, 95% CI = − 0.410, − 0.090; D⁺E⁻: p < 0.001, aMD = − 0.375, 95% CI = − 0.582, − 0.168). Median regression with genotype and stimulus pair as independent variables, and each animal treated as a cluster. b See Additional file 1: Table S7 for statistics. c Accuracy on A>E (non-transitive) and B>D (transitive) during transitive inference test (dotted line indicates chance performance; AE>chance t₍₅₇₎ = 22.44, p < 0.001; BD>chance t₍₅₇₎ = 3.98, p < 0.001). Data points plotted for each mouse, data represent median ± 95% CI. # denotes significant difference between stimulus pairs (^#p < 0.05, ^##p < 0.01, a = lower than A⁺B⁻, b = higher than B⁺C⁻, c = higher than A⁺B^-); * denotes significant difference from chance (50% accuracy) (*p < 0.05, **p < 0.01)