The present paper completes a restudy of the main lava pile in the Deccan flood basalt province (trap) of India. Chenet et al. (2008) reported results from the upper third, and this paper reports the lower two thirds of the 3500-m-thick composite section. The methods employed are the same, i.e., combined use of petrology, volcanology, chemostratigraphy, morphology, K-Ar absolute dating, study of sedimentary alteration horizons, and as the main correlation tool, analysis of detailed paleomagnetic remanence directions. The thickness and volume of the flood basalt province studied in this way are therefore tripled. A total of 169 sites from eight new sections are reported in this paper. Together with the results of Chenet et al. (2008), these data represent in total 70% of the 3500-m combined section of the main Deccan traps province. This lava pile was erupted in some 30 major eruptive periods or single eruptive events (SEE), each with volumes ranging from 1000 to 20,000 km 3 and 41 individual lava units with a typical volume of 1300 km(3). Paleomagnetic analysis shows that some SEEs with thicknesses attaining 200 m were emplaced over distances in excess of 100 km (both likely underestimates, due to outcrop conditions) and up to 800 km. The total time of emission of all combined SEEs could have been (much) less than 10 ka, with most of the time recorded in a very small number of intervening alteration levels marking periods of volcanic quiescence (so-called "big red boles''). The number of boles, thickness of the pulses, and morphology of the traps suggest that eruptive fluxes and volumes were larger in the older formations and slowed down with more and longer quiescence periods in the end. On the basis of geochronologic results published by Chenet et al. (2007) and paleontological results from Keller et al. (2008), we propose that volcanism occurred in three rather short, discrete phases or megapulses, an early one at similar to 67.5 +/- 1 Ma near the C30r/C30n transition and the two largest around 65 +/- 1 Ma, one entirely within C29r just before the K-T boundary, the other shortly afterward spanning the C29r/C29n reversal. We next estimate sulfur dioxide (likely a major agent of environmental stress) amounts and fluxes released by SEEs: they would have ranged from 5 to 100 Gt and 0.1 to 1 Gt/a, respectively, over durations possibly as short as 100 years for each SEE. The chemical input of the Chicxulub impact would have been on the same order as that of a very large single pulse. The impact, therefore, appears as important but incremental, neither the sole nor main cause of the Cretaceous-Tertiary mass extinctions.