Continuous measurements of whole canopy isoprene emissions over an entire growing season are reported from Harvard Forest (42¿32'N, 72¿11'W). Emissions were calculated from the ratio of observed CO2 flux and gradient multiplied by the observed hydrocarbon gradients. In summer 1995, 24-hour average emissions of isoprene from June 1 through October 31 were 32.7¿1010 molecules cm-2 s-1 (mg C m-2 h-1=2.8¿1011 molecules cm-2 s-1), and the mean midday mixing ratio was 4.4 ppbv at 24 m. Isoprene emissions were zero at night, increased through the morning with increasing air temperature and light, reached a peak in the afternoon between the peaks in air temperature and light, and then declined with light. Isoprene emissions were observed over a shorter seasonal period than photosynthetic carbon uptake. Isoprene emission was not detected from young leaves and reached a peak rate (normalized for response to measured light and temperature conditions) 4 weeks after leaf out and 2 weeks after emissions began. The normalized emission rate remained constant for approximately 65 days, then decreased steadily through September and into October. Total isoprene emissions over the growing season (42 kg C ha-1 yr-1) were equal to 2% of the annual net uptake of carbon by the forest. Measured isoprene emissions were higher than the Biogenic Emission Inventory System-II model by at least 40% at midday and showed distinctly different diurnal and seasonal emission patterns. Seasonal adjustment factors (in addition to the light and temperature factors) should be incorporated into future empirical models of isoprene emissions. Comparison of measured isoprene emissions with estimates of anthropogenic volatile organic compound emissions suggests that isoprene is more important for ozone production in much of Massachusetts on hot summer days when the highest ozone events occur. ¿ 1998 American Geophysical Union