Data collection is one of the most important functions provided by wireless sensor networks. In this paper, we study the theoretical limitations of data collection in terms of capacity for a wireless sensor network where n sensors are randomly deployed. We first consider data collection under physical model. We show that the capacity of data collection is in order of Θ(W) where W is the fixed data-rate on individual links. Particularly, we give a simple collection method based on interference blocks and theoretically prove that the method can achieve the optimal capacity in order. If each sensor can aggregate its receiving packets into a single packet to send, the capacity of data collection increases to Θ((n/logn)W) . We then derive a lower bound and an upper bound of data collection capacity under generalized physical model where the data rate depends on the signal to interference plus noise ratio. Data collection is one of the most important functions provided by wireless sensor networks. In this paper, we study the theoretical limitations of data collection in terms of capacity for a wireless sensor network where n sensors are randomly deployed. We first consider data collection under physical model. We show that the capacity of data collection is in order of Θ (W) where W is the fixed data-rate on individual links. Particularly, we give a simple collection method based on interference blocks and theoretically prove that the method can achieve the optimal capacity in order. If each sensor can aggregate its receiving packets into a single packet to send, the capacity of data collection increases to Θ ((n / logn) W). We then derive a lower bound and an upper bound of data collection capacity under generalized physical model where the data rate depends on the signal to interference plus noise ratio.