Sixty-five Norwegian 10th graders used the software Read&Answer 2.0 (Vidal-Abarca et al., 2011) to read five different texts presenting conflicting views on the controversial scientific issue of sun exposure and health. Participants were administered a multiple-choice topic-knowledge measure before and after reading, a word recognition task, and a reading motivation inventory that included two dimensions: Science reading self-efficacy, focusing on readers' beliefs about their capabilities to comprehend what they read in science, and science reading task value, focusing on readers' beliefs about how important, useful, and interesting it is to comprehend science texts. In addition, strategic reading pattern was assessed in terms of the degree of non-linear reading behavior. Multiple regression analysis showed that word recognition skills strongly predicted learning from the texts, as assessed by participants' increase in topic knowledge. However, when multiple-text comprehension indicated by performance on open-ended short-essay questions was the dependent variable, not only word recognition but also strategic reading pattern and science reading self-efficacy emerged as unique predictors when topic knowledge was controlled for. Science reading task value was not related to performance. This study provides new evidence that new literacy competencies needed in a knowledge society, such as synthesizing or integrating across multiple conflicting sources of information, still largely involve word-level, strategic, and motivational processes that may profitably be targeted through systematic instruction.