Most previous studies of preterm birth have considered risk factors in isolation rather than examining the collective impact of multiple candidate determinants. In order to examine the combined impact of a set of behavioral risk factors on the risk of preterm birth, we analyzed data collected for the Pregnancy, Infection, and Nutrition Study on a range of sociodemographic, behavioral, and related factors. Women who received prenatal care at selected clinics in central North Carolina and gave birth in the period 1995-2005 were recruited into a prospective cohort study, with 4,251 women providing the required information on risk factors and pregnancy outcome. A number of demographic and behavioral attributes were modestly associated with preterm birth, with odds ratios of 1.3-1.5, including age >35, African-American ethnicity, height of 63 inches or less, parity 2+, and delivery at the academic medical center. Despite weak associations for individual risk factors, changes in a constellation of behaviors during pregnancy predict substantial shifts in the risk of preterm birth, suggesting a reduction from 8 to 3% preterm among those with a low-risk baseline profile, and a reduction from 18 to 7% preterm among those with a high-risk baseline profile. While inferences are limited by the incomplete range of available predictors, uncertainty regarding whether observed associations are causal, and substantial challenges in changing component behaviors, the possibility of substantial reduction in risk merits more serious consideration of whether behavioral interventions could markedly reduce the risk of preterm birth.