This paper focuses on the analysis of the trade-off existing between achieved performance and observation requirements in a decision making framework based on the cognitive cycle. For that purpose, it presents a general belief based decision making framework that can be particularized for different observation strategies that decide when measurements are carried out to characterize the dynamics of the radio environment. While the concept can be applied to different cognitive radio problems, the paper focuses on the spectrum selection to establish a set of radio links. The considered observation strategies and associated decision making are assessed under different environment conditions in terms of the traffic generation patterns. Results reveal that both the session generation rate and the session duration play a key role to choose the adequate observation strategy that balances the achieved performance and the measurement requirements.