On 28-29 June 2004 a multicellular thunderstorm west of Oklahoma City, Oklahoma, was probed as part of the Thunderstorm Electrification and Lightning Experiment field program. This study makes use of radar observations from the Norman, Oklahoma, polarimetric Weather Surveillance Radar-1988 Doppler, three-dimensional lightning mapping data from the Oklahoma Lightning Mapping Array (LMA), and balloon-borne vector electric field meter (EFM) measurements. The storm had a low flash rate (30 flashes in 40 min). Four charge regions were inferred from a combination of LMA and EFM data. Lower positive charge near 4 km and midlevel negative charge from 4.5 to 6 km MSL (from 0° to -6.5°C) were generated in and adjacent to a vigorous updraft pulse. Further midlevel negative charge from 4.5 to 6 km MSL and upper positive charge from 6 to 8 km (from -6.5° to - 19°C) were generated later in quantity sufficient to initiate lightning as the updraft decayed. A negative screening layer was present near the storm top (8.5 km MSL, ∼ 25°C). Initial lightning flashes were between lower positive and midlevel negative charge and started occurring shortly after a cell began lofting hydrometeors into the mixed phase region, where graupel was formed. A leader from the storm's first flash avoided a region where polarimetric radar suggested wet growth and the resultant absence of noninductive charging of those hydrometeors. Initiation locations of later flashes that propagated into the upper positive charge tracked the descending location of a polarimetric signature of graupel. As the storm decayed, electric fields greater than 160 kV m-1 exceeded the minimum threshold for lightning initiation suggested by the hypothesized runaway breakdown process at 5.5 km MSL, but lightning did not occur. The small spatial extent (≈ 100 m) of the large electric field may not have been sufficient to allow runaway breakdown to fully develop and initiate lightning.