Five years of measurements of PAN and O3 at a rural site in eastern Canada are reported. The measurements were made year-round and only interrupted due to instrumental failures. The data are analyzed by advanced statistical methods to determine temporal trends, both for PAN and O3 individually and with respect to their covariance. Since the data are clearly lognormally distributed, this analysis is performed on the logarithmic transform of the data. PAN shows a statistically significant long-term trend but O3 does not. It is argued that this trend may be the result of a change in the overall hydrocarbon composition of the air, since no trend in NOx emissions appears to have taken place over the same time period in eastern North America. Strong seasonal trends in PAN and O3 are found that are similar in overall pattern but differ in detail. The similarity in shape leads to a significant covariance between PAN and O3. The differences in detail can be interpreted as indicative of well-known atmospheric processes, such as the impact of stratospheric folding occurrences leading to additional O3 with respect to PAN in the spring and reduced photochemical activity and lower temperatures in the winter months, which results in a decoupling of the covariance between O3 and PAN. On a day-to-day basis the covariance between PAN and O3 is strong due to the fact that it is essentially determined by meteorological variation. Episodes of alternatively clean and polluted air are observed, each lasting circa 3 days. Seasonal trends in the diurnal patterns are also evident for both compounds: diurnal variation in the summer is considerable but becomes negligible in the winter. PAN displays a larger diurnal variation than O3 in the summer, which can be explained by their relative dry deposition velocities.