Comics are a popular medium to visually convey narratives. How readers comprehend one of their special features – the combination of text and pictures (“cross-codality”) – is of theoretical and practical interest for narrative comprehension research. Previous experiments (Gernsbacher, 1985; own unpublished conceptual replication) suggest that building mental representations of narratives and integrating cross-codal information could differentially affect the retention of surface information. We used this surface information retention measure to gain a better understanding of the cognitive comprehension processes for cross-codal information in visual narratives (e.g., Comics, picture stories). We tested whether only complete codality-switches trigger changes in surface information retention or if changes are rather guided by the narrative/content relevance of the cross-codal information. In a preregistered online-experiment (N = 148), participants viewed short picture stories in which the critical panel was shown in 4 variations (2x2 within-subjects design): narratively relevant information was shown pictorially or textually, while accompanying narratively irrelevant information could also be shown pictorially or textually, resulting in either completely pictorial, completely textual, or mixed panels. Participants had to comprehend the stories picture by picture and were tested for the recognition of surface information (i.e., left-right orientation) of the panel before the critical event. Results show an increase in surface information recognition only if both relevant and irrelevant information was switched to text. This indicates that the narrative relevance of cross-codal information might play a subordinate role in comparison to completely switching narrative codality. The implications of additional viewing time analyses will be discussed further.