This paper considers an uplink cellular system, in which each base station (BS) is equipped with a large number of antennas to serve multiple single-antenna user equipments (UEs) simultaneously. Uplink training with pilot reusing is adopted to acquire the channel state information (CSI) and maximum ratio combining (MRC) or zero forcing (ZF) reception is used for handling multiuser interference. Leveraging stochastic geometry to model the spatial distribution of UEs, we analyze the statistical distributions of the interferences experienced by a typical uplink: intra-cell interference, inter-cell interference and interference due to pilot contamination. For a practical but still large number of BS antennas, a key observation for MRC reception is that it is the intra-cell interference that accounts for the dominant portion of the total interference. In addition, the interference due to pilot contamination tends to have a much wider distribution range than the inter-cell interference when shadowing is strong, although their mean powers are roughly equal. For ZF reception, on the other hand, we observe a significant reduction of the intra-cell interference compared to MRC reception, while the inter-cell interference and the interference due to pilot contamination remains almost the same, thus demonstrating a substantial superiority over MRC reception.