Chemical modification techniques are widely used in graphite furnace atomic absorption spectrometry (GFAAS). Palladium is a very effective chemical modifier and can be used to stabilize many elements to several hundred degrees higher than the temperatures possible with current methods [1-9]. Of the elements tested, the greatest temperature shifts are achieved for the semi-metallic elements such as As, Se, Te, Bi, Sb, Pb, TI, Ga, Ge and P. Ash temperatures can be raised 400-800 'C higher than current methods allow. Temperature shifts are somewhat less for the transition elements and ash temperatures can be raised 200-500'C. Palladium has no similar effect on elements in Groups I and 11 of the Periodic Table. The change in stability is believed to be due to the formation of an intermetallic species. This improvement in stability permits more efficient removal of matrix constituents during the ash step and vaporization into a hotter environment during the atomize step. Background and interference problems are thus reduced or eliminated. Steps taken to guarantee that palladium is present as the reduced metal as early as possible greatly improve performance of the modifier. The palladium modifier solution can be pre-injected and the graphite tube heated to 1000 'C. Such a method has been used to stabilize mercury [10]. It is assumed that at this temperature palladium metal is present on the graphite surface. The sample can then be introduced. The addition of a reducing agent such as 5% hydrogen in 95% argon, ascorbic acid, or hydroxylamine hydrochloride also appears to guarantee that the palladium is present as the metal early in the temperature program. The use of hydrogen as a reducing agent appears to be the method of choice for a number of reasons. It is cleaner, leaves no residue, and is less subject to contamination. It is also easy to use. A pre-mixed gas of 5% hydrogen in 95% argon can simply be introduced into the furnace. More importantly, the problem encountered with high concentrations of nitric acid is eliminated with the use of hydrogen. Reduced palladium metal allows the retention of the analyte element on the graphite surface until a higher gas phase temperature is achieved. This appears to give many of the analytical advantages normally associated with platform atomization while using the simpler wall atomization technique. Investigations to elucidate the mechanism of palladium chemical modification have been conducted. Also, comparison of palladium modifier methods with current modifier methods in spike recovery studies from difficult matrices was accomplished. Scanning electron micrographs of graphite surfaces with palladium deposits obtained by different reduction methods were obtained to investigate whether the physical form of palladium influenced the modifier behavior. It was found that reduced palladium metal was indeed present on the graphite surface after reduction and that the most effective reduction method was the use of 5% hydrogen in 95% argon with a palladium solution containing 1% glycerol. The scanning electron micrograph of the surface produced under these conditions is shown in figure 1. The palladium particles are considerably smaller and more highly dispersed than those produced by other methods. The average particle diameter is 0.05-0.15 pAm. The smaller particles result in a great increase in palladium surface area. Spike recovery studies from difficult matrices showed that smaller, highly dispersed particles produced improved interference performance. The interference performance of a palladium/ hydrogen/glycerol method for tin was compared with the performance of a commonly used method. The commonly used method requires the use of the platform and chemical modifiers of ammonium dihydrogen phosphate and magnesium nitrate. Wall atomization was used for the palladium method. The results of this study are listed in table 1. While both methods performed well in concentrated HCl, in most instances the palladium gave slightly better interference performance. It was significantly better in overcoming interferences from NaCI and seawater matrices. Other elements have been tested in the same matrices and, in virtually every instance, palladium gave as good as or better interference performance (wall) as did the currently used methods (platform).