Cooperation among mobile users (MUs) in a wireless network can be very useful to reduce the total radiated power necessary to insure the delivery of the information with the desired quality of service. A systematic framework for achieving such a gain consists in making the cooperating nodes act as the antennas of a virtual transmit array, operating according to a distributed space-time coding (DSTC) strategy. However, cooperation implies the allocation of dedicated resources, typically power and time slots, for the exchange of data between source and intermediate nodes (relays). It is then necessary to design the system properly to make possible a final net gain, taking into account all resources involved in the communication. In this paper, we consider regenerative relays and we analyze the effect of intermediate decision errors at the relay nodes. We derive the optimal maximum-likelihood (ML) detector, at the final destination, in case of binary phase-shift keying (BPSK) transmission, and a suboptimal scalar detector, whose bit-error rate (BER) is expressed in (approximate) closed form. Since with DSTC the transmit antennas are not colocated, we show how to allocate the power among source and relay terminals in order to minimize the average BER at the final destination. Finally, we compare alternative cooperation and decoding strategies.