The throughput of slotted ALOHA systems with multipacket message transmissions is evaluated. The conventional strategy of retransmitting the entire message when collisions occur results in a reduction in throughput when compared with the single packet case. However, it is proved analytically that the use of ARQ-type retransmission protocols can provide significant performance improvement. In particular, it is shown that a simple go-back-

(GBN) protocol can offer a modest throughput increase, while a selective reject (SREJ) strategy provides exactly the same throughput as a single packet slotted ALOHA system, irrespective of message length statistics. These results motivate a new scheme (referred to as unslotted selective reject (SREJ) ALOHA) for increasing the throughput of unsynchronized random access channels. It is demonstrated that unslotted SREJ ALOHA with optimized parameters can achieve typical maximum throughputs (after accounting for all overheads) at least 25-40 percent higher than conventional pure ALOHA.