Since the cognitive revolution, scholars of cognition have not stopped questioning about how human beings represent knowledge. Thus, it is postulated that humans construct mental representations and according to the discipline in which these mental representations have been proposed, they receive different names, such as "framework" (Minsky, 1974), "script" (Schank y Abelson, 1987), and "mental model" (Johnson-Laird, 1990), among others. Studies of reading comprehension have not abandoned this assumption that individuals construct representations in their minds. Within this framework, in the late twentieth century, van Dijk and Kintsch (1983) propose that readers (of oral and written texts) processed textual information in three levels of representation: surface code, text-base and situation model. In this article, the focus of interest is this last level because of its impact and acceptance on becoming, along with the other two levels, a "non-controversial assumption" (Graesser, Singer y Trabasso, 1994). More specifically, the discussion will present Kintsch's conception of the situation model (1988, 1998) and will underpin its value for the studies of understanding text. To achieve this goal we focus on the treatment of this mental representation in the construction-integration model (Kintsch, 1988; Kintsch, 1999). Finally, we review, in the light of various experiments (McNamara, 2004; Zwaan y Taylor, 2006; etc.), the psychological validity of situation models. This will reaffirm its value not only for the studies of reading comprehension but also for education and, in general, for cognition (Kintsch, 2004).