In the years to come, German wheat, corn and aggregated cereal yields can be expected to show growing deviations from a linearly increasing trend. This results from the Bayesian Updating approach I apply to agricultural yield data. The updating procedure is carried out on a set of hypotheses on yield development, which are weighted in the light of yield data from 1950 through 2006. All hypotheses share the assumption of a linear yield trend with normally distributed variance of actual data from this trend, but differ in regard to possible future developments. The set of hypotheses allows for both the trend and the variance of data to stay unchanged, increase or decrease by 20 per cent from one period to the next. As a result, yield expectations converge to favor a stable positive linear trend with increasing variance by 1990, at latest. Expectations for the future are stabilized, as the present weight of this hypothesis is at least 99.5 per cent for all crops considered. Impacts of climate change may have contributed to this development in the past, and current knowledge on its effects on agricultural production augments the credibility of these expectations for the future. From the increase in variance, it follows that the risk of losses for agricultural actors increases and should be backed up with the help of hedging or insurances instruments.