The production and downward transport of particulate organic matter (POM) is an important process in the marine carbon cycle affecting the CO${_2}$ exchange between ocean and atmosphere (biological ...pump). Sinking particles export carbon and nutrients from the surface into the deep ocean, and C:N:P:O element ratios of POM determine the relative magnitudes of downward phosphorus, nitrogen and carbon fluxes. Currently, it is common practise to use Redfield ratio C:N:P:O which is constant in space and time for flux estimation and biogeochemical modeling. However, there is evidence that particle compositions underly systematic variations and models using the constant Redfield ratios may underestimate downward carbon fluxes markedly. For the determination of elemental ratios of POM and their impact on the marine carbon cycle we use C/N ratios measured on particles, and we assembled particle data from many different sources into a single data collection for joint evaluation. The dataset contains approximately 9200 single values of C/N ratios, encompassing all major oceans and latitudes, oligotrophic and high productive regions as well as areas of seasonal ice coverage. Analysis of this global dataset shows that C/N ratios are highly variable in space and time, ranging from values below the Redfield ratio (C/N similar to = similar to 6.6) to values greatly exceeding it. There is a systematic and statistically significant trend of C/N ratios increasing with depth by 0.4 units per 1000 similar to m depth. After correcting for the contribution of terrigeneous material C/N ratios of marine POM are also found to increase with depth by about 0.2 units per 1000 similar to m depth. Arguments on how these results from the analysis of POM can be reconciled with previous studies based on dissolved nutrient fields are presented. Depth dependent C/N element ratios should be implemented in biogeochemical models to correctly represent the relative strengths of downward carbon and nitrogen fluxes.