The requirement of ubiquitous information access motivates the development of wireless communication networks. Such networks are expected to support multimedia applications with different traffic characteristics and quality of service (QoS) requirements, similar to wired networks. Hence, in the case of wireless networks it is also important to devise effective resource allocation policies to guarantee a given QoS to different traffic classes. To increase frequencies reuse, a current trend in wireless cellular networks is toward microcellular architectures; this increases handoffs rates, with the possible consequent increase in the probability of connection interruptions. Hence, an important measure of the effectiveness of access control policies for wireless networks is the blocking probability of arriving connection requests in each cell. We consider the problem of optimal access control for a wireless network supporting multiple classes of traffic. In particular, we consider two optimization problems: minimizing any linear function of the blocking probabilities of different classes, and minimizing the blocking probability of one class, with a constraint on the blocking probability of the other class. For the first problem, we prove that the search for the optimal control policy can be limited to policies that base their decisions only on the occupancy levels. This result also implies that hysteresis-based policies are not optimal. For the second problem, we prove that within the class of fixed threshold policies a fractional threshold policy is optimal and provide a simple algorithm to calculate this threshold given the system parameters.