Abstract— In [1], Gupta and Kumar determined the capacity of wireless networks under certain assumptions, among them point-to-point coding, which excludes for example multi-access and broadcast codes. In this paper, we consider essentially the same physical model of a wireless network under a different traffic pattern, namely the relay traffic pattern, but we allow for arbitrarily complex network coding. In our model, there is only one active source/destination pair, while all other nodes assist this transmission. We show code constructions leading to achievable rates and derive upper bounds from the max-flow min-cut theorem. It is shown that lower and upper bounds meet asymptotically as the number of nodes in the network goes to infinity, thus proving that the capacity of the wireless network with n nodes under the relay traffic pattern behaves like log n bits per second. This demonstrates also that network coding is essential: under the point-topoint coding assumption consider...