Forecasting is a critical component of planning, controlling and risk management for construction projects. In order to support effective project execution and control, project managers must be able to make reliable predictions about the final project duration and cost of projects starting virtually from project inception. The objective of this research is to develop probabilistic forecasting models that integrate all relevant information and uncertainties into consistent predictions in a mathematically sound procedure usable in practice. A Bayesian adaptive forecasting framework using S-curves has been developed. The primary advantages of this new approach against conventional methods such as the critical path method and the earned value method are threefold. It is (1) a probabilistic method that provides confidence bounds on predictions; (2) a consistent method that is applicable to both schedule and cost forecasting; and (3) an integrative method that maximizes the value of information - subjective or objective - available from standard construction project management practice. A numerical example is presented to show the adaptive nature of the new method along with the advantages of a probabilistic approach compared to deterministic methods. In the example, the Bayesian model averaging technique is used to combine predictions by different S-curve models and the results indicate that combined predictions outperform individual predictions.