BACKGROUND: Body-borne accelerometers are valuable tools for objectively assessing the frequency, duration and intensity of human physical activity. However, the quality of information received from accelerometers is dependent on the reliability and validity of the devices. PURPOSE: To assess the validity and reliability of the 3DNX™ triaxial accelerometer (BioTel Ltd., UK) in a mechanical setting. METHOD: Nine accelerometers were securely attached to the mounting plate of a Multi Axis Shaker Table (MAST) (MAST-9720, Instron Structural Testing Systems Ltd., UK) using double-sided tape. The units were subjected to eight one-minute stages along each measurement axis (X, Y and Z). The testing schedule spanned a range of frequencies and amplitudes resulting in three conditions at 0.5 g, three at 1.0 g and two at 1.5 g. This schedule was repeated on two occasions. All raw accelerometer data were downloaded and imported into a spreadsheet where 5 s epoch accelerometer counts were averaged over each stage. As each stage was one minute long this was typically the mean of ten values after excluding the first and last epochs of each stage. Intra-unit reliability for each axis was assessed using limits of agreement (LoA) and coefficients of variation (CVintra). Inter-unit reliability was assessed using coefficients of variation (CVinter) and Intraclass correlation coefficients (ICC) which were generated from unit mean values and the mean of all units (3DNX_All). Validity was assessed using simple linear regression. RESULTS: The absolute bias ± 95 % LoA were 0.03 (p>0.05) ± 3.28 counts·5 s-1, -0.23 (p>0.05) ±3.44 counts·5 s-1 and -0.28 (p>0.05) ±4.34 counts·5 s-1 for X-, Y-, and Z-axes, respectively. CVintra for all units ranged from 0.1-0.9 %, 0.0-1.0 % and 0.0-0.9 % for X-, Y-, and Z-axes respectively. CVinter ranged from 0.4-1.3 %, 0.3-0.8 % and 0.4-0.7 % for X-, Y- and Z-axes, respectively. The ICCs were 1.0 for all stages in all axes. The regression between 3DNXTM counts·5 s-1 and acceleration (g) for all axes combined (n=432) was y =211.96x (r2 = 0.99, SEE 6 counts·5 s-1). CONCLUSION: The 3DNXTM is a valid and reliable instrument for measuring accelerations in a mechanical setting.