Deep-sea corals are a new tool in paleocearrography with the potential to provide century long records of deep ocean change at subdecadal resolution. Complicating the reconstruction of past deep-sea temperatures, Mg/Ca and Sr/Ca paleothermometers in corals are also influenced by non-environmental factors, termed vital effects. To determine the magnitude, pattern and mechanism of vital effects we measure detailed collocated Sr/Ca and Mg/Ca ratios, using a combination of micromilling and isotope-dilution ICP-MS across skeletal features in recent samples of Desmophyllum dianthus, a scleractinian coral that grows in the near constant environment of the deep-sea. Sr/Ca variability across skeletal features is less than 5% (2 alpha relative standard deviation) and variability of Sr/Ca within the optically dense central band, composed of small and irregular aragonite crystals, is significantly less than the surrounding skeleton. The mean Sr/Ca of the central band, 10.6 +/- 0.1 mmol/mol (2 alpha standard error), and that of the surrounding skeleton, 10.58 +/- 0.09 mmol/ mol, are statistically similar, and agree well with the inorganic aragonite Sr/Ca-temperature relationship at the temperature of coral growth. In the central band, Mg/Ca is greater than 3 mmol/mol, more than twice that of the surrounding skeleton, a general result observed in the relative Mg/Ca ratios of D. dianthus collected from separate oceanographic locations. This large vital effect corresponds to a similar to 10 degrees C signal, when calibrated via surface coral Mg/Ca-temperature relationships, and has the potential to complicate palcoreconstructions. Outside the central band, Mg/Ca ratios increase with decreasing Sr/Ca. We explain the correlated behavior of Mg/ Ca and Sr/Ca outside the central band by Rayleigh fractionation from a closed pool, an explanation that has been proposed elsewhere, but which is tested in this study by a simple and general relationship. We constrain the initial solution and effective partition coefficients for a Rayleigh process consistent with our accurate Metal/Ca measurements. A process other than Rayleigh fractionation influences Mg in the central band and our data constrain a number of possible mechanisms for the precipitation of this aragonite. Understanding the process affecting tracer behavior during coral biomineralization can help us better interpret paleoproxies in biogenic carbonates and lead to an improved deep-sea paleothermometer. (C) 2007 Elsevier B.V All rights reserved.