While experimental studies in the monkey have shown that motor recovery after partial destruction of the hand motor cortex was based on adjacent motor reorganization, functional MRI (fMRI) studies with isolated primary motor cortical stroke have not yet been reported in humans. Based on experimental data, we designed a study to test if recovery after stroke within primary motor cortex (M1) was associated with reorganization within the surrounding motor cortex, i.e. the motor cortex was able to vicariate. Since motor recovery is time-dependent and might be inflected according to the tested task, the delay after stroke and two motor tasks were included in our design. We examined four patients with one ischaemic stroke limited to M1, and four sex- and age-matched healthy controls in a temporally balanced prospective longitudinal fMRI study over three sessions: <20 days, 4 months and 2 years after stroke. The paradigm included two motor tasks, finger tapping (FT) and finger extension (FE). Distinct patterns of motor activation were observed with time for FT and FE. At the first session, FT-related activation was lateralized in the ipsilateral hemisphere while FE-related activation was contralateral, involving bilateral cerebellar regions for both tasks. From 4 months, skilled motor recovery was associated with contralateral dorsal premotor and sensorimotor cortex and ipsilateral cerebellum motor-related activations, leading to lateralized motor patterns for both tasks. For the left recovered hand, FT and FE-related activations within M1 were more dorsal in patients than in controls. This dorsal shift progressively increased over 2 years, reflecting functional reorganization in the motor cortex adjacent to the lesion. In addition, patients showed a reverse representation of FT and FE within M1, corresponding to a greater dorsal shift for FT than for FE. This functional dissociation might reflect the structural subdivision of M1 with two distinct finger motor representations within M1. Recovery of FT, located within the lesioned depth of the rolandic sulcus in controls, might be related to the re-emergence of a new representation in the intact dorsal M1, while FE, located more dorsally, underwent minor reorganization. This is the first fMRI study of humans presenting with isolated M1 stroke comparable with experimental lesions in animals. Despite the small number of patients, our findings showing the re-emergence of a fingers motor task in the intact dorsal M1 instead of in ventral M1 are consistent with 'vicariation' models of stroke recovery.