Infant mortality rates (IMR) declined dramatically in New York City between 1988-89 and 1992-93. For whites, Hispanics and blacks the City recorded IMR decreases of 27.4%, 24.8% and 22.7% respectively over this period. A change in IMR may result either from changes in birth weight specific death rates(BWSDR) or from shifts in the birth weight distribution (BWD). To explore the relative impacts of these two factors on the recent declines in IMR, we examined two cohorts of linked birth-death certificates, 1988-89 and 1992-93(N=501,324). We estimated separate multinomial logistic regressions for whites, Hispanics and blacks to control for a spectrum of demographic, behavioral, neighborhood and quality of care variables and then used direct and indirect standardization techniques to predict declines in IMR due to shifts in BWSDR versus declines due to shifts in BWD. The model generated predicted IMR declines of -28.8%, -22.9% and -20.0% for whites, Hispanics and blacks respectively due to declines in BWSDR and changes of 6.8%, -0.9 and-2.9% respectively due to shifts in BWD. We conclude that the significant shifts in IMR witnessed in New York City over the past several years were due principally to declines in BWSDR rather than to shifts in BWD. This suggests that the dissemination of technologies, such as artificial surfactant, which improve the survival of low birth weight babies rather than efforts to improve birth weight distributions has had greater recent impact on this important measure of infant health.