
This paper reports measurements of atmospheric neutrino and antineutrino interactions in the MINOS Far Detector, based on 2553 livedays (37.9 ktonyears) of data. A total of 2072 candidate events are observed. These are separated into 905 containedvertex muons and 466 neutrinoinduced rockmuons, both produced by chargedcurrent $\nu_{\mu}$ and $\bar{\nu}_{\mu}$ interactions, and 701 containedvertex showers, composed mainly of chargedcurrent $\nu_{e}$ and $\bar{\nu}_{e}$ interactions and neutralcurrent interactions. The curvature of muon tracks in the magnetic field of the MINOS Far Detector is used to select separate samples of $\nu_{\mu}$ and $\bar{\nu}_{\mu}$ events. The observed ratio of $\bar{\nu}_{\mu}$ to $\nu_{\mu}$ events is compared with the Monte Carlo simulation, giving a double ratio of $R^{data}_{\bar{\nu}/\nu}/R^{MC}_{\bar{\nu}/\nu} = 1.03 \pm 0.08 (stat.) \pm 0.08 (syst.)$. The $\nu_{\mu}$ and $\bar{\nu}_{\mu}$ data are separated into bins of $L/E$ resolution, based on the reconstructed energy and direction of each event, and a maximum likelihood fit to the observed $L/E$ distributions is used to determine the atmospheric neutrino oscillation parameters. This fit returns 90% confidence limits of $\Delta m^{2} = (1.9 \pm 0.4) \times 10^{3} eV^{2}$ and $sin^{2} 2\theta > 0.86$. The fit is extended to incorporate separate $\nu_{\mu}$ and $\bar{\nu}_{\mu}$ oscillation parameters, returning 90% confidence limits of $\Delta m^{2}\Delta \bar{m}^{2} = 0.6^{+2.4}_{0.8} \times 10^{3} eV^{2}$ on the difference between the squaredmass splittings for neutrinos and antineutrinos.