Heisenberg's intuitive notion of uncertainty contains many aspects which need to be distinguished carefully, if one wants give uncertainty a precise quantitative meaning. The textbook relation concerns “preparation uncertainty”: there is no state for which both the momentum distribution and the position distribution are sharply peaked (have low variance). Heisenberg's microscope suggests a conceptually different idea: the sharper a position measurement, the larger the disturbance of momentum it introduces. More general, and also more symmetrical in position and momentum, is the idea of “measurement uncertainty”: in every attempted joint measurement of position and momentum there are necessary deviations of the marginal distributions from the respective canonical observables. I will present a rigorous and quantitative statement, which looks just like the textbook uncertainty relation. However, the Delta-quantities take their meaning from the measurement uncertainty scenario.