Background and aim: The effects of text cohesion and added pictures on acquired knowledge have been heavily studied each in isolation. Furthermore, studies on the effects of specific characteristics of pictures, whether facilitating or hindering, are scarce. Schnotz's ITCP Model (2014) allows to formulate hypotheses regarding the combined effect of text cohesion and presence and level of detail of a picture. This study investigates these hypotheses in the case of children reading scientific texts. Sample: One hundred and one-second-, third-, and fourth-grade pupils with a mean age of 9 years, in the western United States. Methods: Data were collected over three sessions to encompass an understanding of each pupil's knowledge based on prior sessions. Results and conclusions: Results showed a significant increase in pupils' knowledge between pre-test and immediate post-test, but as hypothesized, no significant difference between levels of cohesion. No significant difference between types of pictures was detected. After 1 week, knowledge built with a high cohesive text significantly dropped with low-detail picture, whereas, with high detail, or no picture, there was no significant difference. Results suggested that when participants were given a low-detail picture with a low cohesive text, the integration process of the material was more restricted than with a high cohesive text.