We perform a spectroscopic analysis of 492,450 galaxy spectra from the first two years of observations of the Sloan Digital Sky Survey-III/Baryonic Oscillation Spectroscopic Survey (BOSS) collaboration. This data set has been released in the ninth SDSS data release, the first public data release of BOSS spectra. We show that the typical signal-to-noise ratio of BOSS spectra is sufficient to measure stellar velocity dispersion and emission line fluxes for individual objects. The typical velocity dispersion of a BOSS galaxy is 240 km/s, with an accuracy of better than 30 per cent for 93 per cent of BOSS galaxies. The distribution in velocity dispersion is redshift independent between redshifts 0.15 and 0.7, which reflects the survey design targeting massive galaxies with an approximately uniform mass distribution in this redshift interval. The majority of BOSS galaxies lack detectable emission lines. We analyse the emission line properties and present diagnostic diagrams using the emission lines [OII], Hβ, [OIII], Halpha, and [NII] (detected in about 4 per cent of the galaxies). We show that the emission line properties are strongly redshift dependent and that there is a clear correlation between observed frame colours and emission line properties. Within in the low-z sample around 0.15 < z < 0.3, half of the emission-line galaxies have LINER-like emission line ratios, followed by Seyfert-AGN dominated spectra, and only a small fraction of a few per cent are purely star forming galaxies. AGN and LINER-like objects, instead, are less prevalent in the high-z sample around 0.4 < z < 0.7, where more than half of the emission line objects are star forming. This is a pure selection effect caused by the non-detection of weak Hβ emission lines in the BOSS spectra. Finally, we show that star forming, AGN and emission line free galaxies are well separated in the g - r vs r - i target selection diagram.