A series of silica-supported metal phosphides was prepared by two methods involving the reduction of phosphite (I) or phosphate (A) precursors and was studied for the hydrodeoxygenation (HDO) of 2-methyltetrahydrofuran (2-MTHF). The I method required lower temperature than the A method and resulted in catalysts with higher surface area. The activity was evaluated in a packed-bed reactor on the basis of equal CO chemisorption sites (30 mu mol) loaded in the reactor with comparison made to a commercial Pd/Al2O3 catalyst. At 300 degrees C and 1 atm, the order of activity was Ni2P > WP > MoP > CoP > FeP > Pd/Al2O3. The principal HDO products for the iron group phosphides (Ni2P and COP) were pentane and butane, whereas for the group 6 metal phosphides (MoP and WP), the products were mostly pentenes and pentadienes. For the Pd/Al2O3 and the low-activity FeP/SiO2 catalyst the products were mostly pentenes and C4 mixtures. There were no significant differences in the turnover frequency between materials prepared by the two methods, except possibly for the case of WP. There were likewise no great changes in selectivity toward HDO products at 5% total conversion, except for WP. The differences in the case of WP were attributed to the surface P/W ratio that X-ray photoelectron spectroscopy (XPS) showed to be twice as large for the I method than the A method. Contact-time studies were used to develop reaction networks for the most active catalysts, Ni2P/SiO2 and WP/SiO2. For Ni2P/SiO2 by both methods, the selectivity profiles were similar and could be explained by a rake mechanism with pentenes as primary products, 2-pentanone as a secondary product, and pentane as a final product. In contrast, for WP/SiO2, the selectivity depended greatly on the preparation method, but produced a preponderance of unsaturated compounds. The results could be explained from the surface composition. (C) 2012 Elsevier Inc. All rights reserved.