Narrative texts have been advocated as tools to tackle science learning challenges, and there is even the proposal of a "narrative effect" on learning. We believe it is necessary to examine previous evidence on this effect, as well as to characterize the process of learning through science narrative texts more broadly. In this article, we offer a theoretical review drawing on three frameworks, namely on pedagogical aspects of text learning, linguistic features of texts, and cognitive aspects of text comprehension. Based on that, we analyzed two complementary questions. First, we reviewed 36 studies to ask if science narrative texts can benefit learning and memory outcomes at different educational levels (i.e., the "If" question). We found encouraging evidence for the use of science narrative texts at various educational levels, especially in delayed assessments and longer-lasting interventions. Second, we gathered and linked ideas, hints, and evidence on how the process of learning with science narrative texts takes place, namely on conditions and underlying processes (i.e., the "How" question). There are many features from conditions (texts, learners, activities, wider context) and underlying processes (integration with prior knowledge, affective dispositions, and cognitive abilities) that can help to account for variability in outcomes; yet, ideas and evidence are not always tightly connected. We suggest that education and research should focus on specific narrative effects, that specify with what (texts), with whom (learners), when and where (activities and wider context) these effects occur, as well as "why" (underlying processes). We believe the proposed framing can help both make sense of previous evidence and inform future educational practices and research and provide some recommendations in this regard.