The author evaluates two different ways of estimating reach and frequency of plans for printed media. The first assigns reading probabilities to groups of respondents and calculates reach and frequency by simulation. the second estimates parameters to a model for reach/frequency. It is concluded that, in order to prevent bias, ratings per group must be used as reading probabilities. Nevertheless, in most cases, the estimates are still biased compared with panel data, thus overestimating net ´reach. Models with the same assumptions as with assignments of reading probabilities are presented and estiamtes from such models are shown to be closer to panel data. the problem, however, is to get valid input for such models from readership surveys. Means for this are discussed.