A "Generative-Predicational Model" is proposed and applied to the generation of meanings of simple mathematical word-problems. The model suggests that a fundamental property of cognition is a generative process that takes arguments and that produces results, such as events, answers and inferences. This fundamental property, called predication, generates a task-environment i.e., a problem and its corresponding problem-space i.e., its solution. More precisely, a task-environment is a predication consisting of a written mathematical problem and a writer's life experience. A problem-space is a predication consisting of a learner's problem solving schema and of the meaning that the learner generates for the text.The ease with which relations can be established between a task-environment and a problem-space depends on the problem's "coherence" and "complexity" and the learner's experiences and thought processes. Faceted definitions of task-envirortment and problem-space are used to analyze talk-aloud protocols of fifty Israeli sixth-graders tested with thirty word-problems. The empirical results support the proposed model.