This study investigated the test-retest reliability of chemosensory eventrelated potentials in humans. Olfactory event-related potentials and chemosomatosensory event-related potentials were evaluated in 20 healthy, normosmic subjects. Phenyl ethyl alcohol (PEA, 40% v/v) and H₂S (4 ppm) served as olfactory stimuli whereas CO₂ (60% v/v) was the chemosomatosensory stimulus. Fifteen stimuli of each compound were applied to each nostril. Identical stimulation sequences were used during three test sessions. Sessions 1 and 2 were separated by a mean of 6.8 days; sessions 2 and 3 by 12.5 days. Electroencephalographic recordings were made from Fz, Cz, Pz, C3, and C4. Amplitudes and latencies of P1, N1, P2, N2, and P3 were measured. Pearson's correlation coefficient was calculated to test the test-retest reliability, and the general linear model examined the differences. Most correlations ranged between 0.4 < r < 0.75. Latencies correlated significantly better (P = 0.008) between sessions than amplitudes, even though with CO₂ stimulus amplitudes correlated significantly better than with PEA (P = 0.006) or H₂S (P = 0.003). No differences arose between measurements from different nostrils for any stimulus. Chemosensory event-related potentials show good test-retest reliability. Carbon dioxide amplitudes exhibit better signal-to-noise ratios than PEA or H2S amplitudes. Chemosensory event-related potentials are a clinically valuable objective and are reproducible.