Experimental models of Parkinson's disease have demonstrated abnormal synaptic plasticity in the corticostriatal system, possibly related to the development of levodopa-induced dyskinesias (LID). We tested the hypothesis that LID in Parkinson's disease is associated with aberrant plasticity in the human motor cortex (M1). We employed the paired associative stimulation (PAS) protocol, an experimental intervention involving transcranial magnetic stimulation (TMS) and median nerve stimulation capable of producing long-term potentiation (LTP) like changes in the sensorimotor system in humans. We studied the more affected side of 16 moderately affected patients with Parkinson's disease (9 dyskinetic, 7 non-dyskinetic) and the dominant side of 9 age-matched healthy controls. Motor-evoked potential (MEP) amplitudes and cortical silent period (CSP) duration were measured at baseline before PAS and for up to 60 min (T0, T30 and T60) after PAS in abductor pollicis brevis (APB) and abductor digiti minimi (ADM) muscles. PAS significantly increased MEP size in controls (+74.8% of baseline at T30) but not in patients off medication (T30: +0.07% of baseline in the non-dyskinetic, +27% in the dyskinetic group). Levodopa restored the potentiation of MEP amplitudes by PAS in the non-dyskinetic group (T30: +64.9% of baseline MEP) but not in the dyskinetic group (T30: -9.2% of baseline). PAS prolonged CSP duration in controls. There was a trend towards prolongation of CSP in the non-dyskinetic group off medications but not in the dyskinetic group. Levodopa did not restore CSP prolongation by PAS in the dyskinetic group. Our findings suggest that LTP-like plasticity is deficient in Parkinson's disease off medications and is restored by levodopa in non-dyskinetic but not in dyskinetic patients. Abnormal synaptic plasticity in the motor cortex may play a role in the development of LID.