Analysis of observations and sensitivity experiments with a new three-dimensional global model of stable carbon isotope cycling elucidate processes that control the distribution of delta C-13 of dissolved inorganic carbon (DIC) in the contemporary and preindustrial ocean. Biological fractionation and the sinking of isotopically light delta C-13 organic matter from the surface into the interior ocean leads to low delta C-13(DIC) values at depths and in high latitude surface waters and high values in the upper ocean at low latitudes with maxima in the subtropics. Air-sea gas exchange has two effects. First, it acts to reduce the spatial gradients created by biology. Second, the associated temperature-dependent fractionation tends to increase (decrease) delta C-13(DIC) values of colder (warmer) water, which generates gradients that oppose those arising from biology. Our model results suggest that both effects are similarly important in influencing surface and interior delta C-13(DIC) distributions. However, since air-sea gas exchange is slow in the modern ocean, the biological effect dominates spatial delta C-13(DIC) gradients both in the interior and at the surface, in contrast to conclusions from some previous studies. Calcium carbonate cycling, pH dependency of fractionation during air-sea gas exchange, and kinetic fractionation have minor effects on delta C-13(DIC). Accumulation of isotopically light carbon from anthropogenic fossil fuel burning has decreased the spatial variability of surface and deep delta C-13(DIC) since the industrial revolution in our model simulations. Analysis of a new synthesis of delta C-13(DIC) measurements from years 1990 to 2005 is used to quantify preformed and remineralized contributions as well as the effects of biology and air-sea gas exchange. The model reproduces major features of the observed large-scale distribution of delta C-13(DIC) as well as the individual contributions and effects. Residual misfits are documented and analyzed. Simulated surface and subsurface delta C-13(DIC) are influenced by details of the ecosystem model formulation. For example, inclusion of a simple parameterization of iron limitation of phytoplankton growth rates and temperature-dependent zooplankton grazing rates improves the agreement with delta C-13(DIC) observations and satellite estimates of phytoplankton growth rates and biomass, suggesting that delta C-13 can also be a useful test of ecosystem models.