A model of situation model construction suitable for descriptive text, called the Situational Operator Model, was tested in 2 experiments. It assumes that situation models for descriptions are built incrementally by using basic operations that are triggered by text content. In both experiments, participants read and reread texts that described simple machines. Sentence reading times were predicted by the number of operations along with variables associated with text-level features. The impact of the number of operations predicted reading times more consistently for better readers than for poorer readers and for the second reading as opposed to the first. The presence of a drawing task (Experiment 1) or learning a diagram of the machines (Experiment 2) increased the correlation between the number of operators and reading time for better readers but lowered them for poorer readers (Experiment 1). The results provided initial support for the model, and its relation to other models of comprehension is discussed.