A type of item used frequently in standardized testing involves the recognition of a sentence. Examples of such items are the sentence completion items, used in both the SAT and GRE, and the sentence correction items used in the Test of Standard Written English. Tests that contain such items are constructed by a laborious process which, remarkably, does not involve at any point a detailed analysis of the semantic or syntactic properties of the sentence on which the item is based. This paper provides an initial exploration of the possibility that a mental model of the item solution process may provide indications of how difficult it is to solve a certain item correctly. Some of the cognitive theories of language comprehension are reviewed to identify factors that may affect the level of effort required to solve a sentence-based item. It is a first step towards a test development process that does not rely exclusively on empirical test data analysis and instead views the characteristics of items as a source of psychometric information.