Paranodal axoglial junctions formed by the association of contactin-1, contactin-associated protein 1, and neurofascin-155, play important functions in nerve impulse propagation along myelinated axons. Autoantibodies to contactin-1 and neurofascin-155 define chronic inflammatory demyelinating polyradiculoneuropathy subsets of patients with specific clinical features. These autoantibodies are mostly of the IgG4 isotype, but their pathogenicity has not been proven. Here, we investigated the mechanisms how IgG subclasses to contactin-1 affect conduction. We show that purified anti-contactin-1 IgG1 and IgG4 bind to paranodes. To determine whether these isotypes can pass the paranodal barrier, we incubated isolated sciatic nerves with the purified antibody or performed intraneural injections. We found that IgG4 diffused into the paranodal regions in vitro or after intraneural injections. IgG4 infiltration was slow and progressive. In 24 h, IgG4 accessed the paranode borders near the nodal lumen, and completely fill the paranodal segments by 3 days. By contrast, control IgG, anti-contactin-1 IgG1, or even anti-contactin-associated-protein-2 IgG4 did not pass the paranodal barrier. To determine whether chronic exposure to these antibodies is pathogenic, we passively transferred anti-contactin-1 IgG1 and IgG4 into Lewis rats immunized with P2 peptide. IgG4 to contactin-1, but not IgG1, induced progressive clinical deteriorations combined with gait ataxia. No demyelination, axonal degeneration, or immune infiltration were observed. Instead, these animals presented a selective loss of the paranodal specialization in motor neurons characterized by the disappearance of the contactin-associated protein 1/contactin-1/neurofascin-155 complex at paranodes. Paranode destruction did not affect nodal specialization, but resulted in a moderate node lengthening. The sensory nerves and dorsal root ganglion were not affected in these animals. Electrophysiological examination further supported these results and revealed strong nerve activity loss affecting predominantly small diameter or slow conducting motor axons. These deficits partly matched with those found in patients: proximal motor involvement, gait ataxia, and a demyelinating neuropathy that showed early axonal features. The animal model thus seemed to replicate the early deteriorations in these patients and pointed out that paranodal loss in mature fibres results in conduction defects, but not conduction slowing. Our findings indicate that IgG4 directed against contactin-1 are pathogenic and are reliable biomarkers of a specific subset of chronic inflammatory demyelinating polyneuropathy patients. These antibodies appear to loosen the paranodal barrier, thereby favouring antibody progression and causing paranodal collapse.