The ability to recognize emotional facial expressions relies on brain functional networks, particularly in the right hemisphere, and is impaired in neurodegenerative conditions such as Alzheimer’s disease and Parkinson’s disease. Given that this ability involves brain networks operating within hundreds of milliseconds, we tested the hypothesis of abnormalities in event-related spectral electroencephalographic coherence, with a focus on the right versus left hemispheres in these patients. Event-related theta (4–7 Hz) magnitude-squared coherence was calculated for both intra-hemispheric (frontal–temporal, frontal–parietal, central–temporal, and related pairs) and inter-hemispheric homologous electrode pairs (F3–F4, C3–C4, T7–T8, TP7–TP8, P3–P4, O1–O2). We enrolled 25 patients with amnestic mild cognitive impairment, 15 Parkinson’s disease patients with mild cognitive impairment, 25 Alzheimer’s disease patients with dementia, and 16 Parkinson’s disease patients with dementia, along with 25 healthy elderly as controls. Participants were presented with three different facial expressions (angry, happy, and neutral) 60 times each in a pseudo-random order. Theta (4-7 Hz) spectral coherence was higher in the right than the left hemisphere across all groups, with Parkinson’s disease patients exhibiting the lowest values. Moreover, interhemispheric theta coherence was lower in Parkinson’s disease patients with mild cognitive impairment and dementia compared to the amnestic mild cognitive impairment and healthy elderly groups. These findings indicate that cortical functional connectivity related to emotional facial expression processing is more disrupted in Parkinson’s disease than in Alzheimer’s disease, at both mild cognitive impairment and dementia stages. This disruption affects not only the right hemisphere but also interhemispheric connectivity. Finally, it would occur at theta frequencies.