Fig. 5. Thalamocortical coupling at the end of encoding underlies retrograde amnesia pattern.

a Schematic of network model with ablation of HPC at testing time. b Memory recall accuracy as a function of consolidation time with HPC ablation in the testing phase (as in a) of the protocol shown in Fig. 2b. Two-sided Mann–Whitney U test between accuracy in THL and CTX at consolidation time = 0 (p-value = 7.088721 × 10⁻³) and 24 h (p-value = 2.480842 × 10⁻²). c–f Simulation with weakly coupled THL → CTX. c Mean weight strength of THL → CTX synapses clustered according to engram cell preference at the end of training. Two-sided Kolmogorov–Smirnov test between the distribution of THL → CTX weights among engram cells encoding the same stimulus (i.e., diagonal) and that of THL → CTX weights among engram cells encoding different stimuli (i.e., off-diagonal): p-value = 1.659955 × 10⁻¹⁵¹. d Mean weight strength of THL → CTX synapses clustered according to engram cell preference after 24 h of consolidation. Two-sided Kolmogorov–Smirnov test between the distribution of THL → CTX weights among engram cells encoding the same stimulus (i.e., diagonal) and that of THL → CTX weights among engram cells encoding different stimuli (i.e., off-diagonal): p-value = 3.599719 × 10⁻¹⁰. e Memory recall accuracy as a function of consolidation time with the intact network in the protocol depicted in Fig. 2b. Two-sided Mann–Whitney U test at consolidation time = 0 (HPC vs. CTX p-value = 5.583617 × 10⁻³; HPC vs. THL p-value = 0.920340; THL vs. CTX p-value = 5.583617 × 10⁻³) and 24 h (HPC vs. CTX p-value = 0.193816; HPC vs. THL p-value = 7.290358 × 10⁻³; THL vs. CTX p-value = 5.583617 × 10⁻³). f Memory recall accuracy as a function of consolidation time with HPC ablation in the testing phase (as in a) of the protocol shown in Fig. 2b. Two-sided Mann–Whitney U test at consolidation time = 0 (THL vs. CTX p-value = 5.583617 × 10⁻³) and 24 h (THL vs. CTX p-value = 5.583617 × 10⁻³). g–j Simulation with strongly coupled THL → CTX. g Mean weight strength of THL → CTX synapses clustered according to engram cell preference at the end of training. Two-sided Kolmogorov-Smirnov test between the distribution of THL → CTX weights among engram cells encoding the same stimulus (i.e., diagonal) and that of THL → CTX weights among engram cells encoding different stimuli (i.e., off-diagonal): p-value = 0. h Mean weight strength of THL → CTX synapses clustered according to engram cell preference after 24 h of consolidation. Two-sided Kolmogorov–Smirnov test between the distribution of THL → CTX weights among engram cells encoding the same stimulus (i.e., diagonal) and that of THL → CTX weights among engram cells encoding different stimuli (i.e., off-diagonal): p-value = 0. i Memory recall accuracy as a function of consolidation time with the intact network in the protocol depicted in Fig. 2b. Two-sided Mann–Whitney U test at consolidation time = 0 (HPC vs. CTX p-value = 0.920340; HPC vs. THL p-value = 0.920340; THL vs. CTX p-value = 0.920340) and 24 h (HPC vs. CTX p-value = 1.066227 × 10⁻²; HPC vs. THL p-value = 7.290358 × 10⁻³; THL vs. CTX p-value = 0.177809). j Memory recall accuracy as a function of consolidation time with HPC ablation in the testing phase (as in a) of the protocol shown in Fig. 2b. Two-sided Mann–Whitney U test at consolidation time = 0 (THL vs. CTX p-value = 0.920340) and 24 h (THL vs. CTX p-value = 0.920340). *p-value < 0.05 (see Methods). b–j Color as in Fig. 2a. b, e–f, i, j n = 5 trials. Mean values and 90% confidence intervals shown.