In an effort to determine the in situ production rate of spallation-produced cosmogenic 3He, and evaluate its use as a surface exposure chronometer, we have measured cosmogenic helium contents in a suite of Hawaiian radiocarbon-dated lava flows. The lava flows, ranging in age from 600 to 13,000 years, were collected from Hualalai and Mauna Loa volcanoes on the island of Hawaii. Because cosmic ray surface-exposure dating requires the complete absence of erosion or soil cover, these lava flows were selected specifically for this purpose. The 3He production rate, measured within olivine phenocrysts, was found to vary significantly, ranging from 47 to 150 atoms g-1 yr-1 (normalized to sea level). Although there is considerable scatter in the data, the samples younger than 10,000 years are well-preserved and exposed, and the production rate variations are therefore not related to erosion or soil cover. Data averaged over the past 2000 years indicate a sea-level 3He production rate of 125 ?? 30 atoms g-1 yr-1, which agrees well with previous estimates. The longer record suggests a minimum in sea level normalized 3He production rate between 2000 and 7000 years (55 ?? 15 atoms g-1 yr-1), as compared to samples younger than 2000 years (125 ?? 30 atoms g-1 yr-1), and those between 7000 and 10,000 years (127 ?? 19 atoms g-1 yr-1). The minimum in production rate is similar in age to that which would be produced by variations in geomagnetic field strength, as indicated by archeomagnetic data. However, the production rate variations (a factor of 2.3 ?? 0.8) are poorly determined due to the large uncertainties in the youngest samples and questions of surface preservation for the older samples. Calculations using the atmospheric production model of O'Brien (1979) , and the method of Lal and Peters (1967) , predict smaller production rate variations for similar variation in dipole moment (a factor of 1.15-1.65). Because the production rate variations, archeomagnetic data, and theoretical estimates are not well determined at present, the relationship between dipole moment and production rate will require further study. Precise determination of the production rate is an important uncertainty in the surface-exposure technique, but the data demonstrate that it is feasible to date samples as young as 600 years of age providing that there has been no erosion or soil cover. Therefore, the technique will have important applications for volcanology, glacial geology, geomorphology and archaeology. ?? 1990.