A method for the determination of trace levels of silicon from biological materials by inductively coupled plasma mass spectrometry (ICP-MS) has been developed. The volatility of water-soluble silicon species, hexafluorosilicic acid (H₂SiF₆), and sodium metasilicate (Na₂SiO₃) was investigated by evaporating respective solutions (50 µg/mL silicon) in nitric acid (HNO₃), nitric acid + hydrochloric acid (HNO₃ + HCl), and nitric acid + hydrochloric acid + hydrofluoric acid (HNO₃ + HCl + HF) at 120 °C on a hot-block to near dryness. The loss of silicon from H₂SiF₆ solutions was substantial (>99%) regardless of the digestion medium. Losses were also substantial (>98%) for metasilicate solutions heated in HNO₃ + HCl + HF, while no significant loss occurred in HNO₃ or HNO₃ + HCl. These results show that H₂SiF₆ species were highly volatile and potential losses could confound accuracy at trace level determinations by ICP-MS if digestates prepared in HF are heated to eliminate HF. Among the various matrices comprising major elements, sodium appeared to be effective in reducing silicon loss from H₂SiF₆ solutions. Excess sodium chloride (NaCl) matrix provided better stability, improving silicon recoveries by up to about 80% in evaporated HF digestates of soil and mine waste samples, but losses could not be fully prevented. To safely remove excess acids and circumvent the adverse effects of excess HF (e.g., risk of high Si background signals), a two-step digestion scheme was adopted for the preparation of biological samples containing trace silicon levels. A closed-vessel digestion was performed either in 4 mL of concentrated HNO₃ and 1 mL of concentrated HCl or 4 mL of concentrated HNO₃, 1 mL of concentrated HCl and 1 mL of concentrated HClO₄ on a hot plate at 140 °C. Digestates were then evaporated to incipient dryness at 120 °C to remove the acids. A second closed-vessel digestion was carried out to dissolve silicates in 0.5 mL of concentrated HNO₃ and 0.5 mL of concentrated HF at 130 °C. After digestion, digestates were diluted to 10 mL. The solution containing about 5% HNO₃ and 5% HF was directly analyzed by ICP-MS equipped with an HF-inert sample introduction system. The limit of detection was about 110 µg/L for ²⁸Si when using the Kinetic Energy Discrimination (KED) mode. The method was used to determine silicon in various plant and tissue certified reference materials. Data were acquired for ²⁸Si using KED and standard (STD) modes, and ⁷⁴Ge and ¹⁰³Rh as internal standard elements. There was not any significant difference between the accuracy and precision of the results obtained with ⁷⁴Ge and ¹⁰³Rh within the same measurement mode. Precision, calculated as relative standard deviation for four replicate analyses, varied from 5.3 (tomato leaves) to 21% (peach leaves) for plant and from 2.2 (oyster tissue) to 33% (bovine liver) for tissue SRM/CRMs. Poor precision was attributed to material heterogeneity and the large particle size distribution. An analysis of lung tissue samples from those with occupational exposure to silica dust revealed that tissues possessed substantial levels of water-soluble silicates, but the most silicon was present in the particulate matter fraction.