Background: The thalamus, a relay organ consisting of several nuclei, is shared between the frontocerebellar circuit and the Papez circuit, both particularly affected in Alcohol Use Disorder. Shrinkage of the thalamus is known to be more severe in alcoholics with Korsakoff's syndrome than in those without neurological complications (uncomplicated alcoholics). While thalamic atrophy could thus be a key factor explaining amnesia in Korsakoff’s syndrome, the loci and nature of alterations within the thalamic nuclei in uncomplicated alcoholics and patients with Korsakoff’s syndrome remains unclear. Indeed, the literature from animal and human models is disparate regarding whether the anterior thalamic nuclei, or the mediodorsal nuclei are particularly affected and would be responsible for amnesia Methods: Sixty-two participants (20, healthy controls; 26 uncomplicated alcoholics and 16 patients with Korsakoff’s syndrome) underwent a DTI sequence and T1-weighted MRI. State-of-the-art probabilistic tractography was used to segment the thalamus according to its connections to the prefrontal cortex and cerebellar Cruses I and II for the frontocerebellar circuit's executive loop, the precentral gyrus and cerebellar lobes IV-VI for the frontocerebellar circuit's motor loop, and hippocampus for the Papez circuit. The connectivity and volumes of these parcellations were calculated. Results: Tractography showed that the hippocampus was principally connected to the anterior thalamic nuclei while the prefrontal cortex was principally connected to the mediodorsal nuclei. The fibre pathways connecting these brain regions and their respective thalamic nuclei have also been validated. ANCOVA, with age and gender as covariates, on connectivity measures showed abnormalities in both patient groups for thalamic parcellations connected to the hippocampus only (F(2,57)=12·1; p<0·0001; η2=0·2964; with graded effects of the number of connections from controls to UA to KS). Atrophy, on the other hand, was observed for the prefrontal parcellation in both patient groups and to the same extent compared to controls (F(2,56)= 18·7; p<0·0001; η2= 0·40). For the hippocampus parcellation, atrophy was found in the KS group only (F(2,56)= 5·5; p=0·006; η2=0·170, corrected for multiple comparisons using Bonferroni, p<0.01). Post-hoc Tukey’s test for unequal sample sizes, healthy controls > patients with Korsakoff’s syndrome, p=0·0036). Conclusions: Two different mechanisms seem to affect the thalamus. In the FCC, atrophy of the mediodorsal nuclei may lead the alterations, whereas in the PC, disconnection between the anterior nuclei and hippocampus may be the leading factor. Shrinkage of the anterior nuclei could be specific to KS patients, hence a potential neuroimaging marker of its pathophysiology, or more generally of thalamic amnesia for which KS has historically been used as a model.